Modified approach to arithmetical decoder performance optimization for aerial photography images compression

Abstract

This article is devoted to the development of fast arithmetical encoding algorithm for compressing digital images. Subject of this paper is the research of arithmetical encoder complexity dependency on set of criteria beside the problem input size. The task of this article is to research those encoder compartments, which are the most computationally expensive and optimize their implementations. Aim of this research is the development of fast arithmetical encoder as a part of still image codec for compressing the images coming during distant aquatory objects scanning. New mean algorithm complexity estimation method is proposed as well as optimized arithmetical encoder algorithm based on mentioned method. Theoretical research is conducted, results of which are proved with numerical experiment. Actual set of satellite images of Azov sea aquatory was used. Performance of arithmetical encoder is improved by 7%. New method for mean algorithm complexity assessment is proposed which is based on partitioning of inputs array into equivalence classes. Obtained practical results allows increasing newest image compression techniques performance and using them on mobile computational platforms including those, which are installed on UAVs. Theoretical results of this article expands set of methods for assessing the mean algorithm complexity for those cases when number of steps doesn’t depend on problem’s input size but rather on non-measureable criteria such as memory access pattern to RAM from multiple ALUs.