presentation: reviews and talks on icme 2016disp.ee.ntu.edu.tw/class/main.pdfpresentation: reviews...

Presentation:

Reviews and talks on ICME 2016

Cho-Ying Wu

Disp Lab

Graduate Institute of Communication Engineering National Taiwan University

December 07, 2016

Cho-Ying Wu Disp Lab 1 / 70

ICME is a high-class conference on Multimedia.

Cho-Ying Wu Disp Lab

Introduction

2 / 70

Multimedia ?

Graphics

Speech

Vision

Display

Data Mining

Top conference on multimedia and related

Introduction

3 / 70

CVPR / ICCV, ECCV, ACM MM

Generally, ICME is suitable for MS(30%), PhD(65%) students to

challenge. Recruitments from industrial are also a lot.

Topics this year : ( Video and Vision are the most popular in the trend)

Introduction

4 / 70

Roadmap

1 Impressive Talks (1) Fei-Fei Li, Associate Professor, “A Quest for Visual Intelligence.”Computer

Science Dept. Director, Stanford Artificial Intelligence Lab.

(2) Mathias Wien, “High Efficiency Video Coding – Coding Tools and Specification: HEVC V3 and Coming Developments”

Interesting Works

(1) Gene Cheung, Xian Ming Liu “Graph Signal Processing for Image Compression

and Restoration,” Nii.

(2) Xian Ming Liu, “Random Walk Graph Laplacian based Smoothness Prior for

Soft Decoding of JPEG Images,” HIT.

Cho-Ying Wu Disp Lab 5 / 70

Cho-Ying Wu Disp Lab 6 / 70 Cho-Ying Wu Disp Lab

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Semantic Segmentation

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Fei-Fei Li

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Sparsity Model

Mathias Wien

Gene Cheung (1) Gene Cheung, Xian Ming Liu “Graph Signal Processing for Image Compression

and Restoration,” Nii.

Traditional signal property:

1. Discrete

2. Smooth

3. Band-limited frequency

Ex. DCT ( Approx. of KLT )

X x n kN

Dictionary representation inducing sparse coding

Gene Cheung

Graph Signal : Signal live on graph

Ex. Images: 2D – grid

Research interest:

1. Sampling : how to efficiently acquire signal from graph

2. Representation : Given the graph signal, how to compactly

represent it.

3. Signal Restoration : given the partial or noisy signal, how to

recover the structure.

Gene Cheung

Graph Laplacian :

1. Adjancency Matrix

Remark: Sometimes the original problem is hard to solve, we can

change the original problem (primal problem) into dual problem

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

D: distortion R: bit-rate

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Graph Laplacian :

Gene Cheung

Wei Hu, Gene Cheung, Antonio Ortega, Oscar Au, "Multiresolution Graph Fourier Transform for Compression of Piecewise Smooth Images," IEEE Transactions on Image Processing, vol.24, no.1, pp.419-433, January 2015.

Gene Cheung

Skipped part 1. PWS image coding : Generalized GFT with intra-coding of H.264 2. Lifting implementation : lowering complexity from O(N2) to O(Nlog(N)) 3. Image denoising with sparsity and smoothness prior

Xianming Liu : Soft decoding of JPEG images

Ln:Normalized Laplacian See p.56

presentation: reviews and talks on icme 2016disp.ee.ntu.edu.tw/class/main.pdfpresentation: reviews...

Documents

icom-icme 51st annual conference is...

icme 2015 poster

lafert s.p.a. asynchronous icme s.p.a. motors …

brief intro to apache spark @ stanford icme

willkommen zu...

seasons greetings to everyone in icme & very best wishes for...

paper number icme 214

icme educaon at northwestern - carnegie mellon...

icme prize: kickoff meeting documents/2016 icme... · icme...

c lecomte reviews , publications ,talks , posters and...

prospectus for conference sponsors - 6-10 july 2020 ›...

icme 2009 schedule buet

stikes icme jombang.docx

icme-3 indigenous knowledge panel

icme stanford 20110507 final

catalog icme ecab

program icme 2016 v6 icme_2016_v6.pdf5. icme-112: 2d...

icme final, actas de eduación matematica

cloud-based icme software training

2nd world congress on integrated computational materials...