A study of psychovisual image processing

Abstract

There is now strong evidence for the presence of radial and angular sensitivity in the cells of the human visual cortex. Several psychophysical and physiological investigations have shown that response of visual of cortex is turned to a channel (band-limited portion) of the frequency domain and a large variety of human visual system models have been presented in literature. The representation proposed here has for main objective to respect this human like filter bank. Filter banks are defined in discrete Fourier domain. In order to improve the performances, many image coding schemes have been developed by including some of the psychovisual features. Following this line of direction, the purpose of this paper is to determine the efficiency of an image coding scheme which is entirely based on the psychophysical aspects. we purpose to describe a new image transform specifically designed for an efficient coding while preserving a psychovisual Fourier domain description. Also, proposed image coding scheme, based on HVS, can make us index and retrieve images effectively. An original signal is decomposed into basic signals, which include psychovisual features and subband coding, by using matching pursuit algorithm. The matching pursuit is a signals decomposition algorithm that decomposes input signal into a linear expansion of basic signals. Basic signals are selected from a redundancy dictionary of functions. Since basic signals in the dictionary are chosen greedily in order to match the signal, the signal can be decomposed into a few number of basic signals thereby providing compression of the signal. In this paper, we use 2-D matching pursuit algorithm. We show that proposed image coder is both well-suited for psychovisual quantization and channel adapted coding. Consequently, the whole algorithm is performed in the spatial domain. Finally, The results, obtained with a coding scheme which includes these basic features of the visual system, show that at low pSN...