the - iaria journals

The International Journal on Advances in Telecommunications is published by IARIA.

ISSN: 1942-2601

journals site: http://www.iariajournals.org

contact: [email protected]

Responsibility for the contents rests upon the authors and not upon IARIA, nor on IARIA volunteers,

staff, or contractors.

IARIA is the owner of the publication and of editorial aspects. IARIA reserves the right to update the

content for quality improvements.

Abstracting is permitted with credit to the source. Libraries are permitted to photocopy or print,

providing the reference is mentioned and that the resulting material is made available at no cost.

Reference should mention:

International Journal on Advances in Telecommunications, issn 1942-2601

vol. 12, no. 3 & 4, year 2019, http://www.iariajournals.org/telecommunications/

The copyright for each included paper belongs to the authors. Republishing of same material, by authors

or persons or organizations, is not allowed. Reprint rights can be granted by IARIA or by the authors, and

must include proper reference.

Reference to an article in the journal is as follows:

<Author list>, “<Article title>”

International Journal on Advances in Telecommunications, issn 1942-2601

vol. 12, no. 3 & 4, year 2019, <start page>:<end page> , http://www.iariajournals.org/telecommunications/

IARIA journals are made available for free, proving the appropriate references are made when their

content is used.

Sponsored by IARIA

www.iaria.org

Copyright © 2019 IARIA

International Journal on Advances in Telecommunications

Volume 12, Number 3 & 4, 2019

Editors-in-Chief

Tulin Atmaca, Institut Mines-Telecom/ Telecom SudParis, FranceMarko Jäntti, University of Eastern Finland, Finland

Editorial Advisory Board

Ioannis D. Moscholios, University of Peloponnese, GreeceIlija Basicevic, University of Novi Sad, SerbiaKevin Daimi, University of Detroit Mercy, USAGyörgy Kálmán, Gjøvik University College, NorwayMichael Massoth, University of Applied Sciences - Darmstadt, GermanyMariusz Glabowski, Poznan University of Technology, PolandDragana Krstic, Faculty of Electronic Engineering, University of Nis, SerbiaWolfgang Leister, Norsk Regnesentral, NorwayBernd E. Wolfinger, University of Hamburg, GermanyPrzemyslaw Pochec, University of New Brunswick, CanadaTimothy Pham, Jet Propulsion Laboratory, California Institute of Technology, USAKamal Harb, KFUPM, Saudi ArabiaEugen Borcoci, University "Politehnica" of Bucharest (UPB), RomaniaRichard Li, Huawei Technologies, USA

Editorial Board

Fatma Abdelkefi, High School of Communications of Tunis - SUPCOM, TunisiaSeyed Reza Abdollahi, Brunel University - London, UKHabtamu Abie, Norwegian Computing Center/Norsk Regnesentral-Blindern, NorwayRui L. Aguiar, Universidade de Aveiro, PortugalJavier M. Aguiar Pérez, Universidad de Valladolid, SpainMahdi Aiash, Middlesex University, UKAkbar Sheikh Akbari, Staffordshire University, UKAhmed Akl, Arab Academy for Science and Technology (AAST), EgyptHakiri Akram, LAAS-CNRS, Toulouse University, FranceAnwer Al-Dulaimi, Brunel University, UKMuhammad Ali Imran, University of Surrey, UKMuayad Al-Janabi, University of Technology, Baghdad, IraqJose M. Alcaraz Calero, Hewlett-Packard Research Laboratories, UK / University of Murcia, SpainErick Amador, Intel Mobile Communications, FranceErmeson Andrade, Universidade Federal de Pernambuco (UFPE), BrazilCristian Anghel, University Politehnica of Bucharest, RomaniaRegina B. Araujo, Federal University of Sao Carlos - SP, BrazilPasquale Ardimento, University of Bari, ItalyEzendu Ariwa, London Metropolitan University, UKMiguel Arjona Ramirez, São Paulo University, BrasilRadu Arsinte, Technical University of Cluj-Napoca, RomaniaTulin Atmaca, Institut Mines-Telecom/ Telecom SudParis, FranceMario Ezequiel Augusto, Santa Catarina State University, BrazilMarco Aurelio Spohn, Federal University of Fronteira Sul (UFFS), Brazil

Philip L. Balcaen, University of British Columbia Okanagan - Kelowna, CanadaMarco Baldi, Università Politecnica delle Marche, ItalyIlija Basicevic, University of Novi Sad, SerbiaCarlos Becker Westphall, Federal University of Santa Catarina, BrazilMark Bentum, University of Twente, The NetherlandsDavid Bernstein, Huawei Technologies, Ltd., USAEugen Borcoci, University "Politehnica"of Bucharest (UPB), RomaniaFernando Boronat Seguí, Universidad Politecnica de Valencia, SpainChristos Bouras, University of Patras, GreeceMartin Brandl, Danube University Krems, AustriaJulien Broisin, IRIT, FranceDumitru Burdescu, University of Craiova, RomaniaAndi Buzo, University "Politehnica" of Bucharest (UPB), RomaniaShkelzen Cakaj, Telecom of Kosovo / Prishtina University, KosovoEnzo Alberto Candreva, DEIS-University of Bologna, ItalyRodrigo Capobianco Guido, São Paulo State University, BrazilHakima Chaouchi, Telecom SudParis, FranceSilviu Ciochina, Universitatea Politehnica din Bucuresti, RomaniaJosé Coimbra, Universidade do Algarve, PortugalHugo Coll Ferri, Polytechnic University of Valencia, SpainNoel Crespi, Institut TELECOM SudParis-Evry, FranceLeonardo Dagui de Oliveira, Escola Politécnica da Universidade de São Paulo, BrazilKevin Daimi, University of Detroit Mercy, USAGerard Damm, Alcatel-Lucent, USAFrancescantonio Della Rosa, Tampere University of Technology, FinlandChérif Diallo, Consultant Sécurité des Systèmes d'Information, FranceKlaus Drechsler, Fraunhofer Institute for Computer Graphics Research IGD, GermanyJawad Drissi, Cameron University , USAAntónio Manuel Duarte Nogueira, University of Aveiro / Institute of Telecommunications, PortugalAlban Duverdier, CNES (French Space Agency) Paris, FranceNicholas Evans, EURECOM, FranceFabrizio Falchi, ISTI - CNR, ItalyMário F. S. Ferreira, University of Aveiro, PortugalBruno Filipe Marques, Polytechnic Institute of Viseu, PortugalRobert Forster, Edgemount Solutions, USAJohn-Austen Francisco, Rutgers, the State University of New Jersey, USAKaori Fujinami, Tokyo University of Agriculture and Technology, JapanShauneen Furlong , University of Ottawa, Canada / Liverpool John Moores University, UKEmiliano Garcia-Palacios, ECIT Institute at Queens University Belfast - Belfast, UKAna-Belén García-Hernando, Universidad Politécnica de Madrid, SpainBezalel Gavish, Southern Methodist University, USAChristos K. Georgiadis, University of Macedonia, GreeceMariusz Glabowski, Poznan University of Technology, PolandKatie Goeman, Hogeschool-Universiteit Brussel, BelgiumHock Guan Goh, Universiti Tunku Abdul Rahman, MalaysiaPedro Gonçalves, ESTGA - Universidade de Aveiro, PortugalValerie Gouet-Brunet, Conservatoire National des Arts et Métiers (CNAM), ParisChristos Grecos, University of West of Scotland, UKStefanos Gritzalis, University of the Aegean, GreeceWilliam I. Grosky, University of Michigan-Dearborn, USAVic Grout, Glyndwr University, UKXiang Gui, Massey University, New ZealandHuaqun Guo, Institute for Infocomm Research, A*STAR, Singapore

Song Guo, University of Aizu, JapanKamal Harb, KFUPM, Saudi ArabiaChing-Hsien (Robert) Hsu, Chung Hua University, TaiwanJavier Ibanez-Guzman, Renault S.A., FranceLamiaa Fattouh Ibrahim, King Abdul Aziz University, Saudi ArabiaTheodoros Iliou, University of the Aegean, GreeceMohsen Jahanshahi, Islamic Azad University, IranAntonio Jara, University of Murcia, SpainCarlos Juiz, Universitat de les Illes Balears, SpainAdrian Kacso, Universität Siegen, GermanyGyörgy Kálmán, Gjøvik University College, NorwayEleni Kaplani, University of East Anglia-Norwich Research Park, UKBehrouz Khoshnevis, University of Toronto, CanadaKi Hong Kim, ETRI: Electronics and Telecommunications Research Institute, KoreaAtsushi Koike, Seikei University, JapanOusmane Kone, UPPA - University of Bordeaux, FranceDragana Krstic, University of Nis, SerbiaArchana Kumar, Delhi Institute of Technology & Management, Haryana, IndiaRomain Laborde, University Paul Sabatier (Toulouse III), FranceMassimiliano Laddomada, Texas A&M University-Texarkana, USAWen-Hsing Lai, National Kaohsiung First University of Science and Technology, TaiwanZhihua Lai, Ranplan Wireless Network Design Ltd., UKJong-Hyouk Lee, INRIA, FranceWolfgang Leister, Norsk Regnesentral, NorwayElizabeth I. Leonard, Naval Research Laboratory - Washington DC, USARichard Li, Huawei Technologies, USAJia-Chin Lin, National Central University, TaiwanChi (Harold) Liu, IBM Research - China, ChinaDiogo Lobato Acatauassu Nunes, Federal University of Pará, BrazilAndreas Loeffler, Friedrich-Alexander-University of Erlangen-Nuremberg, GermanyMichael D. Logothetis, University of Patras, GreeceRenata Lopes Rosa, University of São Paulo, BrazilHongli Luo, Indiana University Purdue University Fort Wayne, USAChristian Maciocco, Intel Corporation, USADario Maggiorini, University of Milano, ItalyMaryam Tayefeh Mahmoudi, Research Institute for ICT, IranKrešimir Malarić, University of Zagreb, CroatiaZoubir Mammeri, IRIT - Paul Sabatier University - Toulouse, FranceHerwig Mannaert, University of Antwerp, BelgiumMichael Massoth, University of Applied Sciences - Darmstadt, GermanyAdrian Matei, Orange Romania S.A, part of France Telecom Group, RomaniaNatarajan Meghanathan, Jackson State University, USAEmmanouel T. Michailidis, University of Piraeus, GreeceIoannis D. Moscholios, University of Peloponnese, GreeceDjafar Mynbaev, City University of New York, USAPubudu N. Pathirana, Deakin University, AustraliaChristopher Nguyen, Intel Corp., USALim Nguyen, University of Nebraska-Lincoln, USABrian Niehöfer, TU Dortmund University, GermanySerban Georgica Obreja, University Politehnica Bucharest, RomaniaPeter Orosz, University of Debrecen, HungaryPatrik Österberg, Mid Sweden University, SwedenHarald Øverby, ITEM/NTNU, Norway

Tudor Palade, Technical University of Cluj-Napoca, RomaniaConstantin Paleologu, University Politehnica of Bucharest, RomaniaStelios Papaharalabos, National Observatory of Athens, GreeceGerard Parr, University of Ulster Coleraine, UKLing Pei, Finnish Geodetic Institute, FinlandJun Peng, University of Texas - Pan American, USACathryn Peoples, University of Ulster, UKDionysia Petraki, National Technical University of Athens, GreeceDennis Pfisterer, University of Luebeck, GermanyTimothy Pham, Jet Propulsion Laboratory, California Institute of Technology, USARoger Pierre Fabris Hoefel, Federal University of Rio Grande do Sul (UFRGS), BrazilPrzemyslaw Pochec, University of New Brunswick, CanadaAnastasios Politis, Technological & Educational Institute of Serres, GreeceAdrian Popescu, Blekinge Institute of Technology, SwedenNeeli R. Prasad, Aalborg University, DenmarkDušan Radović, TES Electronic Solutions, Stuttgart, GermanyVictor Ramos, UAM Iztapalapa, MexicoGianluca Reali, Università degli Studi di Perugia, ItalyEric Renault, Telecom SudParis, FranceLeon Reznik, Rochester Institute of Technology, USAJoel Rodrigues, Instituto de Telecomunicações / University of Beira Interior, PortugalDavid Sánchez Rodríguez, University of Las Palmas de Gran Canaria (ULPGC), SpainPanagiotis Sarigiannidis, University of Western Macedonia, GreeceMichael Sauer, Corning Incorporated, USAMarialisa Scatà, University of Catania, ItalyZary Segall, Chair Professor, Royal Institute of Technology, SwedenSergei Semenov, Broadcom, FinlandDimitrios Serpanos, University of Patras and ISI/RC Athena, GreeceAdão Silva, University of Aveiro / Institute of Telecommunications, PortugalPushpendra Bahadur Singh, MindTree Ltd, IndiaMariusz Skrocki, Orange Labs Poland / Telekomunikacja Polska S.A., PolandLeonel Sousa, INESC-ID/IST, TU-Lisbon, PortugalCristian Stanciu, University Politehnica of Bucharest, RomaniaLiana Stanescu, University of Craiova, RomaniaCosmin Stoica Spahiu, University of Craiova, RomaniaYoung-Joo Suh, POSTECH (Pohang University of Science and Technology), KoreaHailong Sun, Beihang University, ChinaJani Suomalainen, VTT Technical Research Centre of Finland, FinlandFatma Tansu, Eastern Mediterranean University, CyprusIoan Toma, STI Innsbruck/University Innsbruck, AustriaBožo Tomas, HT Mostar, Bosnia and HerzegovinaPiotr Tyczka, ITTI Sp. z o.o., PolandJohn Vardakas, University of Patras, GreeceAndreas Veglis, Aristotle University of Thessaloniki, GreeceLuís Veiga, Instituto Superior Técnico / INESC-ID Lisboa, PortugalCalin Vladeanu, "Politehnica" University of Bucharest, RomaniaBenno Volk, ETH Zurich, SwitzerlandKrzysztof Walczak, Poznan University of Economics, PolandKrzysztof Walkowiak, Wroclaw University of Technology, PolandYang Wang, Georgia State University, USAYean-Fu Wen, National Taipei University, Taiwan, R.O.C.Bernd E. Wolfinger, University of Hamburg, GermanyRiaan Wolhuter, Universiteit Stellenbosch University, South Africa

Yulei Wu, Chinese Academy of Sciences, ChinaMudasser F. Wyne, National University, USAGaoxi Xiao, Nanyang Technological University, SingaporeBashir Yahya, University of Versailles, FranceAbdulrahman Yarali, Murray State University, USAMehmet Erkan Yüksel, Istanbul University, TurkeyPooneh Bagheri Zadeh, Staffordshire University, UKGiannis Zaoudis, University of Patras, GreeceLiaoyuan Zeng, University of Electronic Science and Technology of China, ChinaRong Zhao , Detecon International GmbH, GermanyZhiwen Zhu, Communications Research Centre, CanadaMartin Zimmermann, University of Applied Sciences Offenburg, GermanyPiotr Zwierzykowski, Poznan University of Technology, Poland

International Journal on Advances in Telecommunications

Volume 12, Numbers 3 & 4, 2019

CONTENTS

pages: 40 - 50Attempts of Fading Student Support in E-learning -Testing a hypothetical model of minimum support through apilot study-Takeshi Matsuda, Tokyo Metropolitan University, JapanMitsuru Kimoto, Gakken Research Institute for Learning and Education, JapanKen Kuriyama, Gakken Research Institute for Learning and Education, Japan

pages: 51 - 60Noise Level Detection Method for General VideoChinatsu Mori, Kogakuin University, JapanSeiichi Gohshi, Kogakuin University, Japan

��

��

��

��

��

� ��!��"��#�#�

��$��%$��$��

&��

&��'��(��)��

*��+,��

� ��!�#��"��#��#� $#$��"��#��#�

Abstract�� !�� "�� #�� $�� "��%� ��&�� "�� "�� "�"� �� '(� �� %)�$�&� *��"�� +� ��)��,�� -�� .�� "�� #�� "�� "��

Keywords-Online learner support; e-learning; self-regulated learning; fading; minimum supprt model.

(#� (-�'./��(.-

�� 0�� 123��(4'(4��#

�� 0��#4�� 5�� 67�879 �� ::9�999 �� 1;3#&�� # 4� ��-��<�� ;92=� �� 2>�68; �� # �� +� �� +�� ;92?�� <�� #

@��+�� +��A�� <�� 0 �� A��<��+�� #

�� +�� <��#��+�� + �� <�� +��

�� +�� B�� +��C �� <�� # ��+ �� +��#*�� D*')E��#

((#� '�)4��/,.'$4-/F�'F.*�.@'�*�4'��

��A��*')��+��+�� <�� +�� 1:3# @�� 0� �� 5� �� G'�� $�� )�� )��5� +�� +�� <� �� F�� /��<��;99>163# (�� H�� 5 ��<��183#

*') �� + �� 0��#4�� *')�� !�� D@��#2E1>31731=3#(��+�

@��2#� F�� #

40

International Journal on Advances in Telecommunications, vol 12 no 3 & 4, year 2019, http://www.iariajournals.org/telecommunications/

2019, © Copyright by authors, Published under agreement with IARIA - www.iaria.org

��+�� I��#

�� #��+�� A�� *') �� <��#(��+��<�� # �� +��+�� <� �� # '�� *') ��<�� <�� #�0� �� 1?3� ��+��1293��+��+�� +� �� 1223� �� B� ��C12;3#

4�� <� � �� A�� +�� # *� ��<��<�� +�� <��# @�� #

��+�� <�� A�� +�� 5 *')# (� ��+�� A�� <�� <� �� <� �� # 4�� *') �� #��<��+��A�� +�� +# �� +�� <�� #

�� #�#��#,� �� +�� B��C <�� +�� #

(� �� + �� A��<��#(�A�� <�� # �� +�� +�� *')� �� <�� B��C��B��C#�� <��*')�0 �� #

�� +� �� B��C� � �� # @�� +�� + �� *')��+�#

(� �� <�� # ��

�� 12:31263# (�� +�� 1283#�� +�� A�� 0�� +�� <��# �� +��<�� <��A��+��<��12>3#

.� �� 5��1273#4�� <�� A��<��<��H��12=312?3#

�� +� �� <�� 0 �� # �� +� �� <��H�� &��5� �� D)�*E� �� A�� <�� +�� +� +�� #��+�� &�� <�� +�� *')��+�#

�� H��+�#(�*��((� �� <��# �� +��<� ��*��(((��+�� +�� *�� (J# �� +�� #

(((#� ��./*

(� �� +� ��<� ��

�� +� �� +�� +�

�� #

A. Hypothetical Model 4 �� 26 �� +��

�� B��4C� ��$�<� �� +�� +��#.��+�� <� <�� # � ��+�� # �� 4 �� <� ��<�� &��# �� <�� D�� *�� *�� *��E �� 0 �� # *�� <�� )�*��<�� +�� +��#�� <��+��>9�� +�� +��#4�� +� �� +�� <��

41



��# �� 5� �� +� �� +��<�� 5� � ��D��@��#;E#�� +�� #

B. Pilot Program ,�� *') ��

�� <�� #�� # (� �� +� �� .��<�� /��<��;92>� �� #�� <��4 ��#4�� +��+�� 0�� #,� �� *')�� D@��#:E#

�� +�� +��7��=��+�� 0�� < ��# @��# 6 �� #4� �� +��A�� <�� +�� <�� #

@�� +�� +�� 4+�� 0�� ;8�� *� ��<�� ;92># �� +��+�� )��F��A�� 173 �� # @�� 4 ��

�� D�<�� E#

@��;#� �� +#

@��:#� F��#

@��6#� �� #

�4K)�(#� F().�F'.&'4�.��)(-�

/�� ,��F�� +��D27.��# 2?/��#;92>E

)�� 4�$�<��D@�� I @�� E

� �� *�� F��D@��E

*�� .��

�� /��<��

4�� +��

�� <�� # �� +�� <��+�� )�*� �� +�� 0 �� 0 �� D@��# 8E# *�� #

� ��+�� 29 �� # (� �� +�� +� �� +� �� +�� +�� B�� L�� +�� )�* �� +�� C#��<�� +��#

42



@��8#� *��D�� !*��

K��!)��5��<��E#

�� +�� +� �� )�*� ��

�� <��D@��#8E#*��+�� +��+��)�*�� :9�� +��<��#

K�� +��+�� +�� <��+��#� �� +�� <�� #

.�� +��4�� <�� +�� # @�� 4L� �� <�� <�� *')� �� +�� A��+��+��<��#

C. Collected Data *��+��<��

��+� �� <�� M# �� <�� +� �� IA�� A��# �� )�*��5�� +�� A��# �� 5 �� 0� ��+�� A�� +� +�� # �� +�� +��H��A�� 0�� D�� 4 ��0 4 �� K �� A��E# (� ��

�4K)�((#� 4J4()4K)�/4�4

��0��

Profile Questionnaire Access Log

*��

-��

&��

*��

F��F��

4��

)�*

'��

��

.��

�

��

,��0 ��

-��

��

��

F��F��

4��

)�*

��

'��

��

,��

��

��4

/��

* ��

)�*

&��

��+�� +��0�� <��#

(� ��+�� L�� #�� +� �� <�� +� ��+�� #4�� +��+�+��*')<�� <�� #

(J#� '�*�)�*

@�� +� �� +�

��H�� +� ��

�� +��

�� #

A. Summary of Logs �<��N��+�� 5 ��

)�*#(�� <�� # (� �� +� �� <�� D2;=E��<��#��+��?>��+��22��+��;2�� +�� D@��# >E# �� +� ��H�� 297 ��<�� <�� #4�� <��+��;2��+��H�<��0��#

4��<��<��26 �� +�� #��<�� +�2�896�+�� 297#6�� #��0��<�� +�;6;��+�6D�<��NE#�� +� �� +�� +�� +� �� A��+��+��+��+��#

43



�4K)�(((#� )�4'-�'*5).&*.-)�4'-(-&�4-4&��-�*O*��

,� �� 1��

��-��<��D��#�� E 2;= 2896 ;8=

4��-��<�� *�� ?#26 297#6: 2=#6:

'�� 0��J�� 2? ;6; :=

��J�� 6 6 8#(��29>��+��

@��>#� K��+��#

�4K)�(J#� ��**4&��.-��-�*@.'�4�� -�.'

�� Total1 2 3 4 5

F�� *�� 2: :; 2> ;2 2 =:

��*�� 26 ;? 27 2: ; 78

F�� ; 8 6 6 9 28

)��F�� : > 2> ; 2 ;=

��&�� ? ;= 7 ? 2 86

.�� 9 ; 2 9 9 :

Total 62 29; >2 6? 8 ;8=#� ��8+��<��<��

B. Messsages �� <�� <� � ��

�� +�;8=�<��>:#: �� D+�� 0��+��+�� <��<��E# �� <�� +� 2=#6 ��;#6 �� #4�� +��+��<��

��D�<�� E#

4�� +�� <�� +�� +�� #�� +�� +��+�� <��#

4�� +��+�� +�� <� �� +��<��#

.��26��+�� I�� ;7�� +� �� # .� � �� <�� ?; �� 5�� D:8#7P �� E# (� �� +��

�� D67��E� �� +��D;;E#�� #.�� <��<�� #�� +�� <� <�� +��+��#

C. Observed Patterns (� �� L� ��

��+�� <�� 5� �� #4��+��<��Q��+�� #@�� <�� D�� E��<� �� <� �� #

�� + ��<��+�� +� �� <�� 5� �� # �� 2� ��+� �� @��# 7� �� + �� <� ��4# (� �� +�� +�� +��#�� +�� <� �� # �� :6��297��#

�4K)�J#� /(*�'(K��(.-.@).&

2��3� ��3� 4��3-��<�� )�� 67 2�:82 29>

'�� :#6P ?>#>P #�0��+��

@��7#� �� 2��+ ��D:6R297��E#

44



,�� ;� �� <�� <�� 5� �� 2� <�� D@��# =E# �� +�� 69��297��#

�� : �� + �� +�� +�� 2��;�� #4�+�� <� ��<�� <�� D@��#?E#�� :+��A�� 2��;��2;��297��#

@��=#� �� ;��+ ��D69R297��E#

@��?#� �� :��+ ��D2;R297��E#

*�� +�� <�� +�� #(�� +�� 22 ��# ��+��29��+�� +��0 �� ;2 ��#

�� +�� +�� +�� # �� +�� )�*��+�� +�� #.� �� 0�� +�� +�� +� �� )�* <�� #.�� #�� +��#

�<��S��+��<�� +��#�+�� #

@�� #)��+�� +��<�� 0�� <�#��<��29�� #�� 5 �� -��<�� ;2 ��/��<�� =� �� -��<�� ;6�� /��<��2;�� <�� +�� # (� +� �� +�� +�� #��+��22�� # 4� ��+��+��+�� #

4��<��S��<��/��<��22�� 2��:��#(��<�� #

*�� 2��:�<�� L��# �� +�� #�� ;��2+��#�� +�� 2 �� ; �� +��#.��

�4K)�J(#� �(��*�'(�*/(*�'(K��(.-.@)�4'-(-&�OF�

,� 5��

6��"� �� 1�� 1 2 3 Other No MSG

.��<��D��28��E 2: 29 6

:

6

:6

2 29R-��# : = 9 ; 2 26

22 ;9R-��# 6 > 2 2 2 2:

;2 :9R-��# : : ; ; 2 22

2 29R/��# > > : 2 : 2?

22 2?R/��# 8 7 ; 2 2 2>

�� :6 69 2; 29 22 297

45



��+�� : �� + ��+�� +�� +�� <��#��0��<��#

D. Awareness of Students �� 5��+� ��

<� ��A�� #,�� A��+��B(��(+�<��+��<��CD(��<��>E��BCD(��<��;;E#��2:��D;:PE�� <��>��;;�� A��H��#

)�� +�� +�� B� ��5� �� <��<�� C �� B�� C#-��B�� C� B�� <�� 0�� C� �� B�� +��C � �� D�<��TE#

��H��<�� <� �� # �� +� ��+� �� 0��# (� �� +�� +�� #

E. Perspectives of e-mentors '�� 5 �� +��0��

�� A�� +��#(��A�� +� �� <�� B�0 �� <� �� <��C#*�� 0� �� <�� 0� <��#(�+��+�� #4�� <�� <��B�� C �� B�� +C �� + �� A��+�� #

�4K)�J((#� 4-*,�'*�.U��*�(.--4('�KO*��/�-�*,�.

�J4)�4��/�.-(�.'(-&KO� ��-�.'F.*(�(J�)ODN=10E

7��"��D-��<��E

0��DPE

��DPE

1��8�� DPE

.��>D6E 6D299E 6D299E ; D89E

.��;;D2E 2D299E 2D299E 2 D299E

K��>��;;

D8E

6D=9E :D>9E 6D=9E

D@��E

-�� +��

D:E

;D>7E

2D::E

;D>7E

.��+�� 0��!

�� B*�� ( +� +�� +��#C

�� B*�� +��#C

�� <�� 5� �� <�� )�*#

(� �� +� �� + �� <�� <��I��+�� 29 �� :9 ��# ��<��+�� D= 29��E ��4�0 ��<�� <��#

-�0�� 0�� A�� +�+��<��+�� #4�� +�� +�� !

�� <�� +�� +�� 5 ��#

�� (� �� < �� <�� <��#

�� B(�� +��<�<��#C

�� B��(+�� +�� #C

.� �� 0 �� +��+�!

�� B,��(��+�� (��#C

�� B(+�� +� ��<��+��L��#C

@�� : �� <��<��#��+��+��#

4 �� <� �� +�� +�� #�#� �� :��+ ��#

J#� �.-�)�*(.-4-/@��'�,.'$

(� �� +� ��

<��+��

��+��<��#

A. Limitations 4�� +� � ��

�� <�� +� �� +��!

46



�� /� +� �� D26�� E�� D<��+��E#

�� F�� +�� +�� +�� <� �� +��# (� �� +�� +� �� +��+�� +��A�� #

�� 4� � �� +� ��<�� L�+��#�� +�� *')�� #

�� 0�� <��<��+�#@��<�� <�� +� �� +� �� <�� 0�� +� �� <�� <�� # @�� +� �� <�� <��<��#

B. Discussion ,� �<��

��+�� #�+��D�� 2��;E��+�� 0 �� D�� : ��E ��+�� 0 �� +�# 4�� A��2:��<��<�� +��A��#�� :��<�� <�� +� �� +�� *')I�� +��# �� : ��+�� <��<��5 �� <�� +�� 0 �� 1;93#

(� �� <�� <�� :�� # (� �� +�� +�� <� �� #(��<��<�� A�� #

)��+�� +�� <��#

C. Future Work �� <� ��

�� <� �� !

1) Development of management methods and systems that enable e-mentors to work effectively and efficiently

* ��+�� 5+��+�� 0�� +��

�� <� �� +��)�*��<� ��+��<��#,� �� +�� L��+� �� )�* ��#

2) Support in the SRL phases other than the performance phase

(� �� +��+�+�� *')#

3) Methods of fading or removal of the scaffold 4��<�� 4��

�� A��*')<�� <�� # �� <�� +�� +�� #

�� 5 *')� �� A�� #(�� +��+��+�� +��+�� ;�� :��#

4�$-.,)�/&��-�

�� +�� +� �� <� �� B/�� C �� &�� D@��O�� ;927 ;92?E# ,� �� +�� D+++#��#��E��#

'�@�'�-��*

123� �#��#$��$#$��B/�� *� �� )��C� �� (�� <�� <�� .� ��)��4��&��;92?�� <��#

1;3� �� (��Survey Report on the Actual Situation of Specific Service Industry. 1.��34��<�� !�� !RR+++#��#��#� R��R��R��<�H�R�� ;R�;?R ��R�;?�� ;7# ��;92?#92#:2��

1:3� �#��$#*��#*#4��B*�� )��! �� C�� '�� # 2:� ��# 8� #82? 8;=� .��<�� ;926� ��!��#��R29#;:96R��#;926#2:#8#82?

163� �� F�� Recommendation of the European Parliament and of the Council of 18 December 2006 on Key Competences for Lifelong Learning.1.��34��<��!�� !RR��0#�� #��R)�0��*��R)�0��*��#��V��W.�!)!;99>!:?6!9929!992=!��!F/@

183� �� * �� *�� Curriculum Guideline for Junior High School�1.��3 4��<�� !RR+++#��0�#��#� R�� RX��R��R��X

47



��RXX��@��R��R;92?R9?R;>R262:8;;X99;# ��

1>3� K# �# Y��+�� /# �# *�� <�� *�� )�� F��# '�� -�+O��;922#

173� /#�#*��K#�#Y��4��*�� '�� )��! �� '�� 4 ��'��-�+O��;997#

1=3� F# '# F�� B4 �� +�� C��F��'��+��#2>� #:=8 697��;996�(**-!2969 7;>;>

1?3� ,#*��#4��$#-��$#$+��B)�� *� �� <� .� )�� U��H*��C�� *��@�HH��(��(��#;>��#;� #>:7 >6>�4 ��;926��!��#��R29#:28>R��#;>#>:7��

1293� �# �� B/� �� 0� �� <�� ()�* G*��5C� (��(�� 4�� )�� (�� 4�� ;996� # 298; 298:� ��!29#229?R(�4)�#;996#2:8776?

1223� �# O�� O# &�� # �� O# *�� #��+� B*�� '��! F�� '�� *� �� )��K��C �� 27�� (�� (�� 4�� )�� D(�4)� ;927E� (�� *�� ;927� # :79 :7;� ��!��#��R29#229?R(�4)�#;927#=8

12;3� �# �� B.��H�� ! /�� C� �� ;99= ,�� %��4��4�� D44��E��;99=� #8:9= 8:2:�(*K-?7= 2 ==99?6 >8 8

12:3� *# K�� *# F�� B*�� 5 �� +�� <�� <�� <�� C �� '�� /�� #8=��#8� #897Z8;?�.��<��;929��!��#��R29#2997R�226;: 929 ?289 ?

1263� $# )# ��-�� /# �# )�H�� # $�� '# ,# ��0�B*� �� 5 �� 0 ��<�� C��)�� *�� # 28� ��# ;� #28:Z2?2� -��<��;99>��!��#��R29#2;97R�28:;7=9?��289;X2

1283� *#&��K#$��B�� 0�� [�� 5�� <��C�(��*��#67� #:?Z>=�-��<��;92=��!��#��R29#2997R�22;82 92= ?678 H

12>3� �# ,�� @# @�� B@�� ! �� #C)�� (�� # ;2� ��# >� # 76> 78>�/��<�� ;922� ��!��#��R29#292>R�#��#;922#98#992

1273� *# F��<�� '# ��<�� B�� 0 �� !,�� +��+��+��VC��F��#69� ��# 2� # 2Z2;� �� ;998� ��!��#��R29#2;97R�28:;>?=8� 6992X2

12=3� �# �� $# J�� B�0 �� C� (�� *�� # 6>� ��# ;� #2=8Z;97� 4 �� ;92=� ��! ��#��R29#2997R�22;82 927 ?6;= �

12?3� 4#4#�+��J#)+�\#&��,#\�� $#$�� B�� 5�� <�� C�� K�� # =7� #6:> 66?�.��<�� ;92=��!��#��R29#292>R�#��<#;92=#92#9:8

1;93� �#��O# *�� #O�� B)��5 F��<��!��J��<��4��*�� '�� )��C�� )��;99? ,�� )�� &��4��4�� D44��E�.��<��#;99?� #;?27 ;?;2�(*K-?7= 2 ==99?6 7> 2

48



4FF�-/(\

4#� F�� U��(��*��D)��*��2!-�� 6#��0�� E

U2#F��<��<��<��B��C��

��#(��BO��C��A��D-�#;8E��<�

��#

2#(��<��+�� #

;#�� #

:#(��#

6#��+��(+��#

8#(��+��*��/��#

>#(��(+�<��+��<��#

7#�0 ��+��<��<� ��+�� <��#

=#*��#

?#(��<��#

29#(��+��<��#

22#(��(�0 ��#

2;#(�+��+��+��+��(��+��#

2:#(��#

26#(+��+�� #

28#(+�<�� #

2>#(�+��+��#

27#(�� +��#

2=#,��(+�<��#

2?#(�� +��(��#

;9#(��<��#

;2#(�+�� <��+�� #

;;#(��(��L�� <��+��#

;:#��<��A��#

;6#�� +�� <��#

;8#*��+��<�� <��#D@��A��+��#E

(��+��G-�5�� A��+��#

. ��!2#/��;#��:#��6#)��8#�� )��

K#� F�� U�� D)��*��2!-�� 6#��0�� E

U��6#F��<��<�� *��/��

��#

2#� (�� +�� <��#

;#� (+��+��+��+��#

:#� (��#

6#� (�� +��<��#

8#� (��#

>#� (��+��#

7#� (��L�+��+��#

=#� (�5��0 �� <�� <��#

?#� (��L��#

29#� (��+��+��#

22#� (��+�0��+��+��#

2;#� ��+�� #

2:#� (�+��<��+�� #

26#� (+��#

28#� (��+��#

49



U��8#K��0 ��+�� V,�� 0��<��

��29 �� #D�0� ��!(�� 89��#E

50



Noise Level Detection Method for General Video

Chinatsu Mori and Seiichi Gohshi

Kogakuin University1-24-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo, Japan

Email: [email protected]

Abstract—Continuous efforts on TV technology have been carriedout over the years with respect to resolution enhancement.Currently, 4K TV is a standard TV and 8K broadcasting isplanned to begin in December 2018. High resolution in con-junction with low noise is an essential figure of merit in videosystems. Unfortunately, any increase in resolution unavoidablyincreases noise levels. A signal processing method called “noisereduction” or “noise reducer (NR)” is often used to reduce noise.However, accurate noise level is needed when NR method isemployed. Noise level depends mostly on lighting conditions andis estimated by comparing adjacent frame difference. However,the frame difference is generated by moving objects as well asnoise. Therefore, it is essential to determine whether the framedifference is caused by the moving objects or by the noise, whichis a difficult task. Another difficulty arises from the fact thatnoise level detection must be achievable in real-time conditionssince all video systems are required to work in real-time. Thismeans that a complex method could not be used for noise leveldetection. In this paper, two noise level detection algorithms arepresented. The combination of two of them is a concise algorithmable to accurately detect the noise level and work in real-timeconditions.

Keywords–Video noise reducer; 4KTV; 8KTV; Real-time; Non-linear signal processing; Image quality.

I. I NTRODUCTION

A dramatic change in imaging technologies has takenplace in the 21st century. High Definition Television (HDTV)broadcasting started only 20 years ago and at that time HDTVsets were expensive. Today, HDTV is already a part of history.4K TV broadcasting started a couple of years ago and 8Ksatellite broadcasting is started in December 2018. Althoughsignificant advances have been made in video resolution,imaging technologies are based on the same principle, i.e.,the photoelectric effect. Imaging devices primarily compriseof photoelectric cells and the number of electrons generatedby each cell is proportional to the number of photons receivedby the cell. As the resolution increases from HDTV to 4Kand then to 8K, the size of the image cell decreases, i.e., thenumber of photons per image cell is inversely proportional tothe resolution. Therefore, it is necessary to amplify the electricenergy of a video signal at the output of a video camera.

The electrical energy generated by the image cell is ampli-fied by a pre-amplifier for each pixel. An amplifying processalways results in thermal noise called “Gaussian noise.” Thelevel of noise is inversely proportional to the electric energygenerated per cell. This is because fewer photons generatea lower voltage signal that requires amplification to achievethe appropriate voltage level. As HDTV, 4K, and 8K arehigh-resolution systems, the noise level increases because the

size of the image cells becomes smaller due to the high-resolution. The best way to reduce noise in a high-resolutionvideo is to increase the sensitivity of image cells’ photoelectriceffect. However, in order to achieve this, there are technicallimitations, which need to overcome. Even high-end matureHDTV cameras may have pulse noise called “Shot noise”under poor lighting conditions, such as night time shootingor shooting in a dark room.

Noise reducer (NR) is a technology able to reduce noisein video systems by using signal processing techniques. Al-though, a large number of NR algorithms have been reportedmost of them are complex and only compatible with stillimages. The use of such an algorithm in real-time videosystems would cause a video to freeze. In other words,complex NR algorithms are not suitable for use in real-timevideo systems. Another issue is the ability to detect accuratenoise levels in video/image systems before applying noisereducing techniques. In case of real-time video systems, noiselevels should be detected in real-time as well. Adjacent framedifference is a basic method to detect noise levels. However,noise, as well as moving objects, is contained in the frame,which makes the detection of accurate noise levels in a real-time video a difficult task. In this paper, a real-time noise leveldetection method is proposed [1].

This paper is organized as follows. In Section II, relatedworks of NR and noise level detection are explained. In SectionIII, two noise level detection algorithms are proposed. InSection IV, simulation results are presented. In Section V, theadvantages and disadvantages of the algorithms are discussedand the combination of two of them is investigated. Finally, inSection VI, conclusions of this work are presented.

II. RELATED WORKS

Conventional NR uses spatial or temporal digital filter toreduce noise [2]–[6]. Many NR methods are used for stillimages. They are spatial digital filters. Generally, the spatialdigital filters cause image blurring. Although the commonmethod is NR with wavelet transformation [7]–[10], the ap-plication of this method in videos is difficult: because real-time performance is required. Hence, an NR with a recursivetemporal filter [11] is the only practical real-time method usedfor videos. Figure 1 shows the signal flow of the conventionalNR method with the temporal recursive filter. The input is anoisy video frame, and the frame memory stores the previousoutput frame, whereα (0 < α < 1) is the NR parameter.Higher values of theα indicate a higher NR effect. However,it is necessary to know the accurate noise level for the NRto work. Generally, videos comprise a wide variety of contentwith different noise levels. The differences are also caused

51



�� Figure 1. NR with recursive filter

by lighting conditions. In the development of automatic, real-time NR hardware, the NR parameter must be set properly inaccordance with the actual noise level of a video. Although theadjacent frame difference is the basis of noise level detection,the frame difference is the result of noise and the movingareas.

Only a few proposals for noise level detection methodsin videos are available. The wavelet transformation is usedfor the noise level detection [12], [13], but its real-time workapplication is difficult owing to its high processing cost.

The spatial and temporal digital filter is simple and is usedfor noise level estimation with low cost [14], [15]. Gaussiannoise can be detected by applying high-pass filter, such asSobel filter and Laplacian filter. However, these filters detectboth noise and temporal moves of videos: the noise level isoverestimated if the video includes fast and complex moves,such as camera works and object moves.

In the authors’ previous works, a noise level detectionmethod, which uses a bilateral filter has been proposed [16].However, the bilateral filter also comes with a high hardwarecost. A noise level detection algorithm is essential not only forthe real-time function but also the accurate determination ofthe actual noise level. The method that uses the bilateral filterfails to perform when the noise level is high. Therefore, someimprovements are necessary to address these issues.

III. PROPOSEDMETHODS

In this paper, two noise level detection algorithms for highand low noise levels area proposed, and the combination ofthese methods is considered.

A. Noise Level Estimation

A video has three axes, namely, vertical, horizontal, and theframe. The plane that consists of the vertical and horizontalaxes is called spatial, whereas the frame axis is called temporal.By comparing the correlations of spatial and temporal, thespatial correlation is stronger than the temporal. The conven-tional NR [11] uses the temporal characteristic, as does thenoise level detection algorithm. However, the adjacent framedifference is the most effective method to detect the noise level,but it involves two types of signals: frame differences causedby noise and that by moving objects in a video.

Figure 2 illustrates some examples. Figure 2 (a) presentsthe frame of a video [17]. In the sequence, trees and leavesrustle in the wind. Figure 2 (b) shows the frame differencecaused by the trees and leaves. The noise level can be obtainedby the standard deviation of the frame difference values inthe flat areas because the frame difference in the flat areasis created by noise. Thus, separating the flat areas with framedifference caused by noise from the areas with moving objects

(a) Frame of a video sequence

(b) Frame difference of (a)

(c) appropriate areas of (b) for noise level detection

Figure 2. Video frame and frame difference

is necessary. The flat areas shown in Figure 2 (a) are markedwith green in Figure 2 (c).

There are two characteristics of the frame differences forseparating the flat areas and moving areas. The frame differ-ence caused by moving objects has shapes and areas, whereasthat caused by noise is isolated. Moreover, moving objects havelarge frame difference values, whereas noise often generatessmall difference values. Based on these characteristics, weintroduce two NR methodologies.

B. Frame Difference and Threshold Process

As discussed in Section III-A, the frame difference valuescaused by the moving objects are larger, thus, distinguishingthese two using a threshold process is possible. Figure 3 shows

52



��

��

��

� ��

��

��

��

��

��

Figure 3. Frame difference and threshold process

Figure 4. Areas detected using the frame difference and threshold process

the block diagram of the noise level detection with framedifference and threshold processing. The frame difference isdetected using a frame memory and the input frame. In thethreshold processing, only a small frame difference is selected,and its values and pixel numbers are sent to the noise levelcalculation block. In the noise level calculation block, theframe difference values and pixel numbers are accumulated.The average noise level can be measured using these twovalues. Figure 4 shows the candidate of the flat areas usingthe frame difference and threshold process. However, thismethod also incorrectly identifies the frame difference causedby moving objects in the tree areas. The moving objects donot always produce large frame difference values. With theluminance-level difference between the moving objects andthe background, the frame difference values are small and cansometimes generate similar values to those caused by noise.Although the frame difference between the blue sky and thetrees is substantial in the video shown in Figure 2 (b), theframe difference among the tree leaves is minimal and similarto the values caused by noise. The incorrect identification dueto similar magnitudes in change between moving objects andnoise is the problem with this method.

C. Proposed Method 1: Area Filter and Edge Detection

As shown in Figure 4, the frame difference caused by thetree leaves is detected as the flat areas for determination ofthe noise level. Although the moving objects, that is, treesand leaves, result in large frame difference values, some canbe quite similar to the nearby areas, such as white shiningleaves and the blue sky. The shining leaves and the skyproduce small frame difference, such as noise because theyhave similar luminance levels. To prevent this issue, we needto connect these areas and exclude the spaces from analysis.Thus, we introduce the area filter and Canny edge detection[18] illustrated in Figure 5, to improve the noise level accuracy.

��

Figure 5. Proposed method 1

��

(a) Interest pixel as flat area

��

(b) Interest pixel as moving area

Figure 6. Examples of area filter process.

Based on the input to the frame difference detection,Figures 5 and 3 are similarly presented. The frame differenceis distributed into three blocks: the area filter, the Cannyedge detection, and the noise level detection in Figure 5. Thefunction of the area filter is illustrated in Figure 6, and isa symmetric nonlinear type of filter. The center pixel valueis processed with the surrounding pixel values and has twoparameters, the kernel size and the threshold level. The kernelsize is5×5, as shown in Figure 6. The input of the area filteris the frame difference and has positive and negative values.

In the area filter block, the frame difference is processedwith an absolute function to render all values positive. Theabsolute values are identified using the algorithm presentedin Figure 6. The white pixels indicate values exceeding thethreshold level, whereas the black pixels are equal to or lessthan the threshold level. If the number of the surroundingpixels exceeding the threshold level is the majority, the areafilter decides the interest pixels as the moving area, otherwise,it decides the interest pixels as the flat area. As shown inFigure 6 (a), the number of pixels exceeding the threshold is11 (the white bocks), and the number equal to or less thanthe threshold is 14 (the black blocks). In this case, the outputof the center pixel is as the flat area. As shown in Figure 6(b), the number of pixels exceeding the threshold is 14 (theblack blocks) and less than or equal to the threshold is 11(the white blocks). Therefore, the output of the center pixel isas the moving area. By using the following method, we candetect most of the moving areas in Figure 4, but not quite all ofthem. Therefore, we also introduce the Canny edge detection.The Canny edge detection identifies the continuous edges inthe frame difference. These edges are caused by the leaves. Acouple of pixels around the Canny detected edges are obtainedfrom the Canny edge detection block. By using the logical ORon the area filter and edge detection blocks, the appropriateareas for the noise level detection are accurately detected.

D. Proposed Method 2: Isolated Point Removal and MotionCompensation

The proposed method 1 shown in Figure 5 can accuratelydetect the noise level when standard deviation is less than 9.We will discuss the problem in the following section in detail.To address the problem that arises when standard deviation is

53



��

��

��

� ��

��

��

��

��

��

��

��

Figure 7. Proposed method 2-A

��

(a) Interest pixel in moving areas

��

(b) Interest pixel in flat areas

Figure 8. Examples of isolated point removal process

Frame

memory

Video

input -

Noise

level

Frame difference

Isolated

point

removal

Noise level

calculation

Motion

compensation

Figure 9. Proposed method 2-B

higher than 9, we have proposed another method.The signal flow of the proposed method 2-A for a high-

level noise is shown in Figure 7. The frame difference detectionprocess of the input and frame memory blocks in Figure 7 isthe same as that in Figure 5. The frame difference is distributedinto two blocks. The first one is the isolated point removalblock and the other one is the noise level calculation block.As discussed in Section III-A, the frame differences are causedboth by moving objects and noise. Given that noise level canbe detected in flat areas, discriminating the flat areas withnoise from the entire frame is necessary. Generally, the framedifferences caused by noise in flat areas are isolated. Whenisolated point removal is used, the output of the isolated pointremoval block can be the same as the frame difference causedby the moving object, and the noise level can be estimatedusing the areas excluding the detected moving areas.

The isolated point removal process is shown in Figure 8.The center pixel in Figure 8 is the interest pixel. Figure 8(a) shows an example where the interest pixel is the movingarea, and Figure 8 (b) illustrates the noise on the flat area. Theinput is the frame difference. Moreover, the absolute value ofthe frame difference is calculated and is binarized using thethreshold level. The pixels shown in Figure 8 are the resultof the binarization. The black areas are below or equal tothe threshold level, indicating the flat area. Meanwhile, thewhite areas are higher than the threshold level, which arecandidates similar to the moving areas or the noise on the flatareas. Using only the flat areas is necessary for the noise levelestimation. Thus, in the isolated point removal process, thecandidate pixels in the white areas are removed if the pixelis isolated and identified as the noise on the flat area. Theparameter of the pixel size of the noise is used and the pixelsize is set to 5 pixels, as shown in Figure 8. As presented inFigure 8 (a), the pixel size of the white area contains 7 pixels,

(a) Proposed method 1

(b) Proposed method 2-A

Figure 10. Areas for noise level estimation

which is larger than 5. The process identifies the area to be themoving area. As shown in Figure 8 (b), the pixel of the whitearea contains 4 pixels, which is less than 5. In this case, thepixel is determined to represent the noise, and it is removed.

Many frame differences are present in the frames. Thesedifferences have larger values when a video includes cameraworks, such as panning and tilting. However, the thresholdprocess cannot detect the frame difference accurately for thenoise level detection. Thus, we also introduce a block-basedmotion compensation to detect and reduce moving areas inthe frame difference. The proposed method 2-A with motioncompensation (method 2-B) is shown in Figure 9. The processof the motion compensation block; the frame is partitionedinto blocks of pixels, and each pixel of a block is shiftedto the position of the predicted block via the motion vector.This process is common in the discussions of video codingtechnologies, such as MPEG-2, MPEG-4, and HEVC. Further-more, we verify and discuss the performance of the motioncompensation in the following sections

The comparison of the areas for noise level estimation withthe proposed methods 1 and 2-A is shown in Figure 10. Thegreen areas show the flat areas that are used to calculate noiselevel, and the black areas show the estimated moving areas.Comparing the proposed method 1 and 2-A, they use differentareas to calculate the nose level, thus, combining two methodsit is able to estimate wide range of noise levels.

54



(a) Sequence 1 (b) Sequence 2 (c) Sequence 3

(d) Sequence 4 (e) Sequence 5

Figure 11. Test sequences

IV. EXPERIMENT

Simulation experiment was conducted to verify the perfor-mance of the proposed methods. Different levels of noise wereadded to video sequences, and the accuracy of the estimatednoise level determined by each method was compared.

A. Test Sequences

Noise levels in general videos were estimated using theframe difference (Section 3.1), the proposed method 1 (Section3.3), and the proposed methods 2-A and 2-B (Section 3.4). Thefive HDTV (1, 920 × 1, 080) video sequences [17] shown inFigure 11 were used in this experiment. Sequence 1 containsthe train moves from the back to the front while the camerapans. In Sequence 2 the woman stands at the same positionwhile only the background moves. The background is insetwith chroma key. Sequence 3 is a pop music show video.Sequence 3 is called confetti and contains flash. Sequence4 looks similar video sequence as Sequence 2. The womanstands at the same position while only the background zoomsback with chroma key. Sequence 5 shows the teams enter themarch and the camera tilts vertically. All sequences includedmoving objects and various camera actions, such as panningand tilting. Gaussian noise with different standard deviations(1, 3, 5, 7, 9, 11, 13, and 15) was added to the videos.

B. Experimental Results

The experimental results are shown in Figures 12 and13. Figures 12 (a)-(e) show the results for sequences 1-5respectively. The figures show the estimated standard deviationfor each level of added noise. The x-axis is the standarddeviation of the noise added to the test sequence, and the y-axisis the estimated standard deviation of the noise in the sequence.The marks show the median values of the estimated standarddeviations. If the estimated noise level is correct, the resulthas the same value as the added noise standard deviation, i.e.,y = x. The bars indicate the minimum to maximum rangeof the estimated noise standard deviation, which shows thevariation of the results in the sequence.

In Figures 12 (a)-(e), the results for the frame difference

method are overestimated and demonstrate large variance. Theestimated results for the proposed method 1 are the mostaccurate and have the smallest dispersion of results. However,the estimation is not possible with the noise standard deviationexceeding 9 because there are few or no appropriate areas forcalculating noise standard deviation. The proposed methods2-A and 2-B returned fewer errors and demonstrate more con-sistent estimated results than the frame difference. However,large errors tend to occur when the noise standard deviationis less than 3. A comparison of the results for the proposedmethods 2-A and 2-B, with and without motion compensation,demonstrates that motion compensation is effective in certaincases. However, it increases the cost significantly because areal-time motion compensation requires large hardware.

Figure 13 shows the estimated noise standard deviation forall frames of sequence 1 (Figure 12 (a)). Figures 13 (a) and(b) show the estimation results for the proposed methods 2-A and 2-B when the noise level is larger than 9. Here thex-axis is the frame number, and the y-axis is the estimatedstandard deviation of the noise in the frame. The resultsbecome constant if the noise level estimation is correct.

In sequence 1, the train is moving with camera panningfrom 0 to 150 frames, then the panning stops. The traincontinues to move during frames 150 to 420. There is nomotion in frames 420-450. As shown in Figures 13 (a) and(b), the effect of motion on the estimation result is negligible,and the results become constant.

The results for all test sequences are shown in Figures 15-19 respectively. Here the x-axis is the frame number, and they-axis is the estimated standard deviation of that. All the resultsshow the same tendensies. The frame difference does not showthe accurate standard deviation and it changes dramatically de-pending on the moving objects. The proposed method 1 givesalmost accurate standard deviation when the noise standarddeviation is lower than 9. However, the proposed method 1does not work when the standard deviation exceeds 9. Theproposed method 2-A produces good results when the standarddeviation is higher than7 except for Sequence 5. Sequence5 has both moving detailed objects and the camera tilting

55



�

�

��

��

��

��

��

��

��

��

��

��

��

�

(a) Sequence 1

�

�

��

��

��

��

��

��

��

��

��

�

(b) Sequence 2

�

�

��

��

��

��

��

��

��

��

��

��

�

(c) Sequence 3

�

�

��

��

��

��

��

��

��

��

�

(d) Sequence 4

�

��

��

��

��

��

��

��

��

��

��

�

(e) Sequence 5

Figure 12. Results of estimated noise standard deviation. (a) - (e) show the results for sequences 1-5, respectively. The estimated results of all frames of thevideo sequence are accumulated. The marks show the median values of the estimated standard deviations. The bars indicate the maximum and minimum values.

56



�

�

�

�

�

��

��

��

��

��

� ��

��

��

� ��

(a) Proposed method 2-A

�

�

�

�

�

��

��

��

��

��

� ��

��

��

� ��

(b) Proposed method 2-B

Figure 13. Results of estimated noise standard deviation in time axis for sequence 1 using the proposed methods (a) 2-A and (b) 2-B

(a) Proposed method 1, standard deviation 3 (b) Proposed method 2-A, standard deviation 3

(c) Proposed method 1, standard deviation 7 (d) Proposed method 2-A, standard deviation 7

Figure 14. Areas of calculated noise standard deviations for one frame of sequence 1

movement. Two different motions generate frame differenceswith different properties. A new method should be added toimprove performance. This is a future work.

V. D ISCUSSION

Comparisons of the areas for noise estimation using theproposed method 1, and the proposed method 2-A are shownin Figure 14. The estimated noise areas for sequence 1 with

added Gaussian noise are shown in Figures 14 (a)-(b) (standarddeviation 3) and Figures 14 (c)-(d) (standard deviation 7).Here, the white areas are estimated moving areas; thus, onlythe black areas are used for noise estimation. When comparingFigure 14 (a) with Figure 14 (b), and Figure 14 (c) with Figure14 (d), the moving areas estimated using the proposed method1 are thick; however, there are few areas for noise estimationwhen the noise level is high. Since the proposed method 1 fully

57



� ��

� ��

��

��

��

�

�

��

�

�

��

��

(a) Frame difference

��

��

��

�

�

�

�

�

�

� ��

(b) Proposed method 1

��

��

��

��

��

�

�

�

��

��

�

�

(c) Proposed method 2-A

Figure 15. Results of estimated noise standard deviation in time axis for sequence 1 using the frame difference (a), the proposed method 1 (b), and theproposed method 2-A (c)

� ��

� ��

��

��

��

�

�

�

��

��


��

��

��

�

�

�

�

�

�

� ��


��

��

��

�

�

�

�

�

�

� ��



� ��

� ��

��

��

��

�

��

�

��

�

�

� � � ��


��

��

��

�

�

�

�

�

�

� ��


��

��

��

�

�

�

�

�

�

� ��



58



� ��

� ��

��

��

��

�

�

�

�

�

�

� ��


��

��

��

�

�

�

�

�

�

� ��


��

��

��

�

�

�

�

�

�

� ��



� ��

� ��

��

��

��

�

��

�

�

�

�

�

�

��

� ��


��

��

��

�

�

�

�

�

�

� ��


��

��

��

�

�

�

�

�

�

� ��



eliminates moving areas, the noise level estimation becomesaccurate. However, the noise estimation does not work withhigh level noise due to few or no available estimation areas.In contrast, the estimated moving areas using the proposedmethod 2-A are thin; therefore, the moving areas of the framewith high level noise are detectable.

As described in Section IV-B, the proposed method 1 candetect low level noise accurately when the standard deviationis less than 9. However, this method requires improvement todetect high level noise when the standard deviation is higherthan 9. In contrast the proposed methods 2-A and 2-B candetect high level noise when the standard deviation is 5 ormore. Therefore, we propose combining the proposed methods1 and 2-A, i.e., when the detected noise level is less than 9,the proposed method 1 is appropriate and when the detectednoise level is equal to or higher than 9, the proposed method2-A is appropriate. Moving compensation can improve noisedetection accurately; however, it requires significantly moreexpensive hardware.

The proposed method 2-A shows higher noise level thanthe actual noise level. Even if a noise level higher than theactual level is used for the noise reducer, it has little effecton the image quality. When the higher noise level than theactual level is input to the noise reducer the noise reducerworks stronger. However, at this noise level: (σ > 9), thereis considerable visual degradation and the effect is difficult toperceive.

In our experiment the proposed method 2-A did not showthe good result in Sequence 5. Sequence 5 has mainly threemotion factors. They are vertical camera tilting, marchingpeople, and others including noise. The global motion vectorcan cover the vertical camera tilting. Although the marchingpeople move almost the same direction, each person movesdifferently. The motion compensation can cover marching peo-ples’movement as a group. However, the motion compensationcannot cope with each person’s movement. If local motionvectors are used, the accuracy of the motion detection wouldbe improved. It would improve precision of the noise level.

59



However, it requires significantly more expensive hardware andit also increases cost.

VI. CONCLUSION

In this paper, real-time noise level detection algorithms forvideos were proposed. They consist of two ideas because itis difficult to cope with the wide range noise level detection.The proposed method is a combination of the low noise leveldetection method (σ < 9) and the high noise level detection(σ ≤ 9). The simulation results demonstrate that the bestresults can be realized by combining two methods. Althoughfurther research is required, the proposed method showedsatisfied results. In future, we intend to develop a way to switchbetween methods automatically and to control NR using theproposed methods. Ultimately, we hope to develop real-timenoise reduction hardware that controls noise level parametersautomatically.

REFERENCES

[1] C. Mori and S. Gohshi, “Real-time noise level detection for generalvideo,” in CONTENT 2019, The Eleventh International Conference onCreative Content Technologies, 2019, pp. 1–6.

[2] M. Kazubek, “Wavelet domain image denoising by thresholding andwiener filtering,” IEEE Signal Processing Letters, vol. 10, no. 11, pp.324–326, 2003.

[3] A. Pizurica, V. Zlokolica, and W. Philips, “Noise reduction in videosequences using wavelet-domain and temporal filtering,” inWaveletApplications in Industrial Processing, vol. 5266. International Societyfor Optics and Photonics, 2004, pp. 48–60.

[4] N.-X. Lian, V. Zagorodnov, and Y.-P. Tan, “Video denoising usingvector estimation of wavelet coefficients,” in2006 IEEE InternationalSymposium on Circuits and Systems. IEEE, 2006, pp. 2673–2676.

[5] I. W. Selesnick and K. Y. Li, “Video denoising using 2d and 3d dual-tree complex wavelet transforms,” inWavelets: Applications in Signaland Image Processing X, vol. 5207. International Society for Opticsand Photonics, 2003, pp. 607–619.

[6] A. Pizurica, V. Zlokolica, and W. Philips, “Combined wavelet domainand temporal video denoising,” inProceedings of the IEEE Conferenceon Advanced Video and Signal Based Surveillance, 2003.IEEE, 2003,pp. 334–341.

[7] N. Gupta, M. Swamy, and E. I. Plotkin, “Low-complexity video noisereduction in wavelet domain,” inIEEE 6th Workshop on MultimediaSignal Processing, 2004.IEEE, 2004, pp. 239–242.

[8] R. O. Mahmoud, M. T. Faheem, and A. Sarhan, “Comparison betweendiscrete wavelet transform and dual-tree complex wavelet transform invideo sequences using wavelet-domain,”INFOS2008, pp. 20–27, 2008.

[9] L. Jovanov, A. Pizurica, S. Schulte, P. Schelkens, A. Munteanu,E. Kerre, and W. Philips, “Combined wavelet-domain and motion-compensated video denoising based on video codec motion estimationmethods,”IEEE transactions on circuits and systems for video technol-ogy, vol. 19, no. 3, pp. 417–421, 2009.

[10] F. Luisier, T. Blu, and M. Unser, “Sure-let for orthonormal wavelet-domain video denoising,”IEEE Transactions on Circuits and Systemsfor Video Technology, vol. 20, no. 6, pp. 913–919, 2010.

[11] J. C. Brailean, R. P. Kleihorst, S. Efstratiadis, A. K. Katsaggelos, andR. L. Lagendijk, “Noise reduction filters for dynamic image sequences:A review,” Proceedings of the IEEE, vol. 83, no. 9, pp. 1272–1292,1995.

[12] V. Zlokolica, A. Pizurica, and W. Philips, “Noise estimation for videoprocessing based on spatio-temporal gradients,”IEEE Signal ProcessingLetters, vol. 13, no. 6, pp. 337–340, 2006.

[13] V. Kamble and K. Bhurchandi, “Noise estimation and quality assess-ment of gaussian noise corrupted images,”IOP Conference Series:Materials Science and Engineering, vol. 331, no. 1, pp. 1–10, 2018.

[14] M. Ghazal, A. Amer, and A. Ghrayeb, “A real-time technique for spatio–temporal video noise estimation,”IEEE Transactions on Circuits andSystems for Video Technology, vol. 17, no. 12, pp. 1690–1699, 2007.

[15] S.-M. Yang and S.-C. Tai, “A fast and reliable algorithm for video noiseestimation based on spatio-temporal sobel gradients,” inInternationalConference on Electrical, Control and Computer Engineering 2011(InECCE). IEEE, 2011, pp. 191–195.

[16] K. Miyamae and S. Gohshi, “Noise level detection in general video,” in2018 International Workshop on Advanced Image Technology (IWAIT),Jan 2018, pp. 1–4.

[17] “Ite/arib hi-vision test sequence 2nd edition,” 2009.

[18] M. J. B. Wilhelm Burger,Principles of Digital Image Processing:Fundamental Techniques. Springer, 2009, pp. 144–145.

60



www.iariajournals.org

International Journal On Advances in Intelligent Systems

issn: 1942-2679

International Journal On Advances in Internet Technology

issn: 1942-2652

International Journal On Advances in Life Sciences

issn: 1942-2660

International Journal On Advances in Networks and Services

issn: 1942-2644

International Journal On Advances in Security

issn: 1942-2636

International Journal On Advances in Software

issn: 1942-2628

International Journal On Advances in Systems and Measurements

issn: 1942-261x

International Journal On Advances in Telecommunications

issn: 1942-2601

the - iaria journals

Documents