The AVQ Algorithm: Watermarking and Compression Performances

摘要

In this paper we review the Adaptive Vector Quantization algorithm for lossy image compression, introduced by Constantinescu and Storer. AVQ combines the potentiality of a dictionary-based algorithm to process input in single-pass with the potentiality of Vector Quantization to approximate data. We discuss an open-source implementation and report the achieved results by this implementation with different size of the dictionary. Subsequently, we consider the problem of the copyright protection in multimedia contents, by focusing our attention on the Digital Watermarking. In addition we describe an approach for this algorithm that permits to improve the robustness of digital invisible watermarks. The proposed approach consists of spreading the watermark into the image during the compression process. We assume that the compression algorithm is aware of the positions of the watermarks: when the algorithm identifies the block containing the watermark, then this block is encoded in loss less mode and is spread all over the image.