Artificial intelligence

Abstract: The paper is devoted to the problem of construction of polynomial separate surfaces for the task of two-class images classification. It is proposed an iteration method that provides to get the coefficients of separate hyper surfaces based on taking into account of peculiarity of the location of learning examples on the boundary between classes. A new algorithm for reducing block artifact structure that occurs in images processed by JPEG algorithm with high compression ratio was proposed. This algorithm combines diadic wavelet transform and the method of optimal interpolation. Its main advantage – simplicity and easy implementation, due to the absence of any threshold techniques in the development process. The proposed method improves the visual quality and peak signal – noise ratio (PSNR).

Keywords:

References: