Given few set of images, this project intelligently generates entirely new possible images using Reduced Singular Value Decomposition(SVD). The code implementation is in matlab.
Keywords : Singular Value Decomposition (SVD), Matlab
Input: Hey program, last night somebody broke into store A. We have image of person with mask from CCTV camera. In the morning, we have got a suspect in our custody and here is his image. Now, can you please say if these two images are of the same person or not?
Output: yes/no.
For the purpose of simplicity, I will just use 3 gray scale images.
- Given image I1, I2 and I3, create data matrix A where each column of represents single image data.
- Apply reduced singular value decomposition(SVD) to obtain basis U that spans the data of image.
- Randomly generate coefficients c1, c2 and c3. Create new linear combination L = c1 * m1 + c2 * u2 + c3 * u3, where u1, u2 and u3 are column of U. The L is the new image data.
- You can create any number of possible images by repeating process 3.
matlab.m is the Source code implemeantion. image1.jpg, image2.jpg, image3.jpg are input images under different light illuminations. simulated_image1.jpg, simulated_image2.jpg,….simulated_image6.jpg are output images of the program. reduced_svd.pdf is slide with SVD algorithm.
Acknowledgements Ths project is named rajSVD to reconize by eldest brother Raj Kumar Lama who supported me in all possible way for my study and SVD algorithm being heart of this implementation. I would also like to acknowledge Dr. Madhusudhanan Balasubramanian, The University of memphis, who made me this algorithm crystal clear.