Home – Home – Jornals – Avtometriya 2023 number 2

A method of adaptive quantization of discrete wavelet transform coefficients in the JPEG2000 standard, based on the algorithm of search and identification of small image structures, which are distinguished by human vision, is proposed. The method of image block classification by structure features and the identification algorithm of coefficients of block transformation in sub-bands of the discrete wavelet transform are described. A description of the adaptive quantization algorithm and recommendations on tuning the quantizer parameters for high image sharpness are given. Experimental estimates of the efficiency of compression of photorealistic images with different detail are provided, which allow conclusions about the advantages of the new method in comparison with analogues to be drawn and recommendations for its further development in intelligent video systems to be given.