Visible to the public HVS-based local analysis of denoising efficiency for DCT-based filters

TitleHVS-based local analysis of denoising efficiency for DCT-based filters
Publication TypeConference Paper
Year of Publication2015
AuthorsRubel, O., Ponomarenko, N., Lukin, V., Astola, J., Egiazarian, K.
Conference Name2015 Second International Scientific-Practical Conference Problems of Infocommunications Science and Technology (PIC S T)
KeywordsBM3D filter, communication system, DCT-based filter, DCT-based filters, denoising efficiency HVS-based local analysis, discrete cosine transform, discrete cosine transforms, Filtering, human vision system, HVS-metrics, image denoising, Image edge detection, image filtering, image visual quality improvement, local analysis, Measurement, multimedia system, noise reduction, noise removal, pre-filtering stage, pubcrawl170111, visualization

Images acquired and processed in communication and multimedia systems are often noisy. Thus, pre-filtering is a typical stage to remove noise. At this stage, a special attention has to be paid to image visual quality. This paper analyzes denoising efficiency from the viewpoint of visual quality improvement using metrics that take into account human vision system (HVS). Specific features of the paper consist in, first, considering filters based on discrete cosine transform (DCT) and, second, analyzing the filter performance locally. Such an analysis is possible due to the structure and peculiarities of the metric PSNR-HVS-M. It is shown that a more advanced DCT-based filter BM3D outperforms a simpler (and faster) conventional DCT-based filter in locally active regions, i.e., neighborhoods of edges and small-sized objects. This conclusions allows accelerating BM3D filter and can be used in further improvement of the analyzed denoising techniques.

Citation Keyrubel_hvs-based_2015