Fast progressive reconstruction of images in DCT domain

Progressive image transmission (PIT) using the discrete cosine transform (DCT) is considered. An approach to fast progressive reconstruction (FPR) of DCT images is presented. The approach is based on the observation that typical input matrices to IDCT are sparse in PIT. By exploiting the sparsity of a matrix, nonzero elements in the matrix can be processed separately and efficiently. The approach allows any scanning patterns including the zig-zag scan and those incorporating human visual sensitivity. This flexibility distinguishes it from recent approaches to FPR of DCT images.