Two Parallel Strategies for Real-time Spatial Video Denoising for Multi-core Processors

Abstract

Video denoising is usually a time consuming process especially for large video files. With the advancement of the processor technology, it is possible to perform video denoising in real-time on multi-core processors. In this paper, we study parallel techniques for denoising real-time video on multi-core processor which work on both shared memory model and distributed memory model. We investigate two approaches: a block approach, which assigns a group of threads to each block of video frames; and a distributor approach, which uses one thread to distribute the frame data to each thread. Our experiments focus on the image denoising technique based on the total variation but the approach can be integrated with other image denoising algorithm like discrete wavelet transform (DWT) or diffusion technique. We found that by using the distributor strategy, we can achieve speedup which is 1.27 times faster than the block strategy and the video frame rate can be increased by 7.43%. Moreover, we also apply the prefetching technique which further enhance frame rate by 22.02% and frame rate control to stabilize frame rate and retain the original video length during denoising and playing in real-time. Our method also has good denoised quality which is better than previous work in [1] in average case. General Terms Algorithms, Framework, Image Processing, Video Processing, High Performance Computing, Parallel Computing Keywords video denoising, parallel computing, OpenMP, ROF model, total variation.