For this problem use the images 'a_g.pgm' and 'b_g.pgm'. For both images do the following using MATLAB:
A. Quantize both images uniformly to 8 representative levels. Evaluate the mean squared error between original and quantized. Submit both original and quantized images.
B. Generate random noise uniformly distributed over [-J,J], where d is the step size for quantization in the previous part, and add this noise to the original images. Now quantize these 'dithered' images uniformly to 8 representative levels. Evaluate the mean squared error between dithered and quantized. Submit both dithered and quantized images.
C. In which case, part (a) or (b) did you get higher MSE? Similarly, in which case, did you observe subjectively better image quality? Explain your answers briefly, covering the following points:
i. Would it make sense to dither certain types of images? If so, then which types?
ii. What are the implications in the perspective of compression?
iii. Is MSE a good measure of image quality?
Note:
Link to 'a_g.pgm' and 'b_g.pgm' files:
https://drive.google.com/drive/folders/1tYkwb0qzlr2IyNI27ncysrmPDLQWEVlC?usp=sharing
