| In recent years, the telecommunications network, radio and television networks andcomputer communication networks mutual penetration, compatible with each other, andgradually integrated into a unified information and communication network. In this case,processing high flow of streaming media services, such as video calling, mobile TV suchhigh bandwidth requirements of the occasion, on some low processing power device,suchas PC and cellphone is high challenge. In view of this situation, the transmission of videonot only need to meet the bandwidth limitations, but also requires a higher visual quality.The encoding bit rate should be as low as possible under the conditions of qualityassurance in order to meet the network restrictions.According to the traditional Graded Quantization model, the frame can be dividedinto ROI and non-ROI in terms of the region of the eye interest in order to improve thesubjective quality of the video with different quantitative parameters (QP) for thecorrespondingly specific region. However, the model can not reflect Human VisualSystem (HVS) well without considering the internal character of ROI. Therefore, for somelow-bit rates coding conditions such as Desktop Video, Handheld terminal with thesubject of human face, we first proposed a face detection method which using skin colourand Gaussian Model. Just-Noticeable-Distortion (JND) model shows that the boundaryarea can conceal more distortion than the smooth region, thus, the boundary in the ROIregion (human face) can be detected for instance, the eye, nose, mouth, etc.Then wepresents a method for Graded Quantization based on ROI and JND, that is,we devide oneframe into background,boundary,ROI_level1,ROI_level2, then creat a four levelQuantization model to guide the quantization of different regions. Compared with thetraditional Graded Quantization modal, the results of the experiment demonstrated that forthe application of the low bit video, the method could improve the subjective video qualityat the same bit rate.MCFI (motion-compensated frame interpolation) technology has been more andmore favored by commercial and consumer electronics manufacturers in recent years. Thisis due to the MCFI improve spatial resolution by inserting additional frames. Accordingly,if we drop some frames adaptively before encoding and then use MCFI to reconstruct thediscarded frame after decoding. This can greatly reduce the amount of encoded data, so asto achieve the purpose of reducing the coding rate. In this paper, we address the problemsof unreliable motion vectors that cause visual artifacts but cannot be detected by high residual energy or bidirectional prediction difference in motion compensated frameinterpolation. A correlation-based motion vector processing method is proposed to detectand correct those unreliable motion vectors by explicitly considering motion vectorcorrelation in the motion vector reliability classifcation, motion vector correction, andframe interpolation stages. Since our method gradually corrects unreliable motion vectorsbased on their reliability, we can effectively discover the areas where no motion is reliableto be used, such as occlusions and deformed structures. We also propose an adaptiveframe interpolation scheme for the occlusion areas based on the analysis of theirsurrounding motion distribution. As a result, the interpolated frames using the proposedscheme have clearer structure edges and ghost artifacts are also greatly reduced.Experimental results show that our interpolated results can reduce31.84%average bit rate,and to ensure a good subjective video quality. |
PDF Full Text Request