ACCELERATION AND EFFICIENCY ANALYSIS OF DCT-BASED FILTERING
Владимир Васильевич Лукин, Sergey Krivenko, Olena Kolganova, М. К. Чобану · Telecommunications and Radio Engineering · 2013
Ways to accelerate image denoising based on discrete cosine transform are considered. A main emphasis is on data processing in partially overlapping blocks of size 8×8 pixels. Two thresholding operations are studied. A simple local adaptation procedure is proposed and its efficiency is analyzed. Comparison of filtering efficiency is carried out using several quality metrics, both standard and based on peculiarities of human visual system. It is demonstrated that, compared to full-overlapping denoising, loss in filtering efficiency can be appropriately small if locally adaptive processing with half block size overlapping and combined thresholding is applied.