Skip to Main Content
A new approach is proposed to solve the blocking artifact problem that results from block transform-based image compression at very low bit-rates. In our approach, we attempt to recover the quantization noise for the transform coefficients by adding uniform random noise to them. This results in a decompressed image that is corrupted by noise of the AWGN type, rather than a blocky image. This recasting of the blocking artifact problem into the domain of AWGN reduction allows any algorithm that has been developed for image denoising to now be a candidate solution for the blocky image problem. In this paper we present initial experiments for this novel approach, using a recently reported denoising algorithm. Results indicate that the algorithm offers both objective and subjective improvements to the blocky image. Additionally, the approach may be generalized to solve artifact problems in other transform domain-based image compression methods.