Analysis-based adaptive wavelet filter selection in lossy image coding schemes

A wide variety of sophisticated wavelet-based lossy image coding algorithms have been developed over the past few years. All of these well-known image coding algorithms use a specific wavelet filter built into the algorithm. Recently we have shown that for better results, rather than using a fixed filter, the most "appropriate" wavelet filter should be chosen adaptively to match the image being coded. In this paper, we present an analysis-based approach to accomplish this in the context of lossy image compression, thereby further improving the performance of a coder. Log energy entropy, variance and range of the wavelet coefficients of the combined subbands are important features in determining the "appropriate" wavelet filter to be used in compressing an image so that the best PSNR can be obtained. There is no need to explicitly code the image using different wavelet filters to determine the best candidate.