Image Restoration for Compression Artifacts by an Associative Memory

Image compression often leads to undesired artifacts. This paper shows an image restoration approach to correct for these artifacts by an associative memory. Investigations for modeling the coder-decoder process by a linear system are described. A compensating system consisting of a neural network based associative memory has been developed to process the degraded result after compression/decompression. For an effective implementation of this method, the image contents must be considered. This way, shortcomings of the mentioned image compression algorithms can be reduced. Key-Words: Image Restoration, Compression Artifacts, Associative Memory, Neural Network, JPEG