Preprocessing method for region-of-interest video call compression

This paper presents a new region-of-interest(ROI) based compression scheme for video call applications in a preprocessing manner. In our application, ROI is the human region which is detected by a human/background segmentation method, and perceptually less important details are removed by applying the edge-preserving spatio-temporal filter to the background regions. In order to obtain the accurate ROI rapidly, we develop a block-wise segmentation scheme and then ROI is decided stochastically. Since our method removes details in the background regions except the edges, we can achieve higher compression rates without sacrificing subjective qualities. Moreover, our method can be applied to any kind of codecs because it works as a pre-processing step. Experimental results show that our method gives accurate ROI and yields similar peak signal to noise ratio (PSNR) in the ROI while achieving much higher compression ratio than the conventional compression methods.