Neural network based image coding

In this paper various soft computing algorithms have been applied for image coding. One of the main common methods to compress images is to code them through vector quantization (VQ) techniques. The principle of the VQ techniques is simple. At first, the image is split into square blocks of pixels, for example 4×4 or 8×8; each block is considered as a vector in a 16- or 64-dimensional space, respectively. Second, a limited number of vectors (codewords) in this space is selected in order to approximate as much as possible the distribution of the initial vectors extracted from the image. Neural network based schemes such as MLP based non-orthogonal transform, nonlinear predictive coding, K-L transform and Kohonen SOM have been selected to achieve this goal. In this sense, this work presents the state of the art in image coding using neural networks.