Transcript
Page 1: Progress  In  Image Registration

Progress In

Image Registration

Page 2: Progress  In  Image Registration

Why Registration

• In computer vision, sets of data acquired by sampling the same scene or object at different times, or from different perspectives, will be in different coordinate systems.

• Image registration is the process of transforming the different sets of data into one coordinate system.

• Registration is necessary in order to be able to compare or integrate the data obtained from different measurements

Page 3: Progress  In  Image Registration

                                                                      

      

                                                                            

                                                                                                               

Example : Two Images From a Mojave Desert Sequence

Page 4: Progress  In  Image Registration

Types of Registration

• Feature Based : Identifies some landmarks, lines, curves, points of high/low intensities and maps them.

• Area Based : looks at the structure of the image as a whole using correlation metrics, Fourier transforms etc.

Page 5: Progress  In  Image Registration

We Use Area Based Reg.

• “In multi-cellular biological images, there are several many different points with similar values of intensity at different cells”

R. Araiza et al. 3-D Image Registration Using Fast Fourier Transformation: Potential Applications to Geoinformatics and Bioinformatics.

Page 6: Progress  In  Image Registration

The Algorithm

Page 7: Progress  In  Image Registration

R. Araiza et al. 3-D Image Registration Using Fast Fourier Transformation: Potential Applications to Geoinformatics and Bioinformatics.

Determining Shift

Page 8: Progress  In  Image Registration

Determining Rotation

• Compute the second order moments of the images :

xdxIxxM ikjijk

)(

• Compare the orientations of the largest eigenvectors of the matrices formed by the second order moments.

Page 9: Progress  In  Image Registration

Determining Scale

• Just divide the magnitudes of the Fourier transforms.

Page 10: Progress  In  Image Registration

Current Status

• A working Code for determining the shift, rotation and scale in 2D images.

(Courtesy : Prof. Bajaj)

• We have assembled an experiment on AVS to check the quality of output of this code.

Page 11: Progress  In  Image Registration

Things To Do

• Subject the 2D code to more tests.

• Extending the code to cater to 3D Images.

• Receive datasets from MDA and run the 3D code on them.

Page 12: Progress  In  Image Registration

Thank You


Top Related