Texture analysis and synthesis using steerable pyramid decomposition for video coding

摘要

Natural textures like grass, water, sand etc. contain large amount of visual details that consume a lot of bitrate while compressing it, but in fact such details are irrelevant for the human vision. Therefore, exploiting human perception properties could provide better compression of such data. In this paper we propose two novel schemes for identifying and processing static and dynamic textures based on 2D Dual Tree Complex Wavelet Transform and Steerable Pyramid Transform. The use of Steerable Filters is motivated by the fact that the frequency and directional selectivity of the Human Visual System can be well modelled by the Steerable Filters.