The problem with thresholding the DCT coefficients of an image to compress it, is that the locations of the chosen coefficients are arbitrary and difficult to code. Consider then using a mask W(k, ℓ)

of the same dimension as the DCT array fC_x(k, ℓ)g. Now the location of the chosen DCT coefficients is known. Consider the cameraman image, compute its DCT coefficients that form a matrix C_x and let

Ĉx = WC_x

be the DCT coefficients of a compressed image. Show the original and the compressed images, and the supports of the original and the masked DCT coefficients. 

Transcribed Image Text:

S1 1