Sparse image compressionIn sparse image compression, we are given the pixel representation of a ‘‘target’’ image, as a vector in . We would like to represent the image as a linear combination of ‘‘basic’’ images , , where the matrix is called the dictionary. Thus, we would like to find coefficients , such that or, more compactly, . Assuming the above equation has solutions, we are interested in finding one solution with the largest number of zeros. This is (empirically) achieved with solving a minimum-norm problem of the form Once a sparse representation of the image is found, we can (say) send the image over the internet by sending only the (few) non-zero coefficients and their corresponding indices . The recipient of these coefficients can reconstruct the image via . This approach assumes that the dictionary is available to both source and recipient. |