Restoration of Images Compressed by Hybrid Compression, based on Discrete Cosine Transform and Vector Quantization, over a Binary Symmetric Channel

Abstract

The H.264 digital video compression standard (encoded files are MP4 files, but they can also be AVI or MKV) had a lot of attention and was successful for a long time due to the many features and techniques used to compress a video. Technically, a video represents a sequence of images, for the H.264 standard, the compression is based on reducing the redundancy in the same frame (an image), which is known as “intra prediction”. This prediction generates modes used to rebuild the image (decompression) when we apply the same idea again but between the frames, we called it as inter prediction. The inter prediction operation also generates other modes. In terms of compression or transmission, predictions mean dependence on primary information, and it is estimated based on primary data, rather than sending all information. In this paper, we employ these prediction modes in the case of intra prediction to correct a single image compressed by vector quantization. This type of compression makes the restoration process very difficult as it produces block-shaped errors and no information about the original block is removed. The results in experimental studies section show a high restoration and correction ratio based on the PSNR as an evaluation parameter. For BER=0.01, we achieved a restoration average of about 0.6 dB, which produces an image with approximately the same PSNR as the original. The main contribution of our method, is that we open a new concept of non- filtering image correction techniques and improve on this correction method, by introducing new prediction modes, to correct the image.