Search for a command to run...
This article is devoted to solving an urgent scientific and applied problem — increasing the efficiency of compression and ensuring the integrity of video information resources under conditions of limited channel bandwidth and the influence of interference. The introductory part substantiates the need for developing new approaches to image segmentation, as traditional methods (JPEG, H.264/AVC, HEVC) and modern formats (JPEG XL) have limited ability to adapt to complex structural features and high vulnerability to channel errors. It is determined that the loss of even a small part of the encoded stream in existing systems leads to disruptions in the decoding process of entire sections due to the lack of structural localization mechanisms. The theoretical section of the research focuses on the transition from the standard ARGB model to the XYB color space, which better matches the characteristics of human perception and avoids ignoring important color components (specifically red). An approach for independent processing of each layer of the color model is proposed, which provides a more efficient division into regions with homogeneous properties. The key contribution of the work is the development of a method for constructing dynamic apertures for asymmetric numeral systems (ANS). Instead of static partitioning, flexible segmentation is introduced, which takes into account local color and structural parameters of the image by dynamically changing the aperture "tube" parameters. The results section presents a comparative analysis of standard 8x8 partitioning, simple aperture partitioning, and the proposed adaptive method. Threshold values for aperture size (from 5 to 100 elements) were experimentally established, at which it is advisable to perform dynamic adaptation of the reference value. This allowed for the optimization of computational costs: encoding time was reduced from minutes to a few seconds, which is crucial for real-time systems. The obtained results confirm that the use of adaptive apertures together with strict value approximation allows for increasing the efficiency of compression using ANS by almost 5 times compared to base approaches, from 228,453 to 40,796 bytes, while maintaining the clarity of significant objects and high noise immunity.