A hybrid image coding technique using a noncausal stochastic model

This paper presents a technique for data compression of a grey-level still image. The image is considered to be a two-dimensional homogeneous random field that satisfies a noncausal stochastic difference equation driven by a white noise field. The difference equation model is isotropic and characterized by the two parameters of the variance and the correlation length of an image only. This model is simple and can easily fit various types of images. Based on the above model, we develop a hybrid algorithm which combines a one-dimensional cosine transformation and a one-dimensional prediction. The algorithm is applied to a real image to check its validity.