Statistical distribution of DCT coefficients and their application to an adaptive compression algorithm
Thierry Eude, Hocine Cherifi, Richard Grisel · 2002
The paper presents a technique for image compression using the Discrete Cosine Transform (DCT) method. In the Joint Photographic Expert Group norm (JPEG), the image is usually compressed using an "universal" quantization matrix. We propose a technique which employs an appropriate distribution model of the DCT coefficients to deduce the quantization matrix from a set of training images. This technique compared to the classical JPEG gave no blocking effect at the same compression rate.>