Dissertation > Industrial Technology > Automation technology,computer technology > Computing technology,computer technology > Computer applications > Information processing (information processing) > Pattern Recognition and devices > Image recognition device

Study on Spatial Error Concealment Algorithm Based on Detail Optimization

Author ZhuLinLin
Tutor ZhaoYan
School Jilin University
Course Communication and Information System
Keywords Airspace error concealment Half-pixel Hessian matrix Stereo video Disparity vector Pixel matching
CLC TP391.41
Type Master's thesis
Year 2011
Downloads 24
Quotes 0
Download Dissertation

Put into use with the 3G network, video applications on the network has been rapid development, has penetrated into every aspect of people's lives. Video data in the video stream, to be transmitted through a wired or wireless network, due to interference, noise, and channel congestion and many other reasons, it will lead to packet loss occurs during transmission, resulting substantial decline in the quality of the receiving end a reconstructed video. The damaged area of ??the reconstructed image to obtain the perception of acceptable video quality is most challenging in the multimedia communication. Error concealment technology refers to the video signal in the time domain or spatial correlation, the signal is lost or damaged, the reconstruction of a technology is divided into airspace, time domain and frequency domain three. This article describes the use of the airspace of the spatial correlation of video information error concealment technique. The first frame (I frame) of the video sequence are intra-coded, and itself has no correlation to the time between the other frames, so in the reconstruction can not use the time-domain redundancy information, the only use of airspace concealment techniques reconstructing lost information. If the bad the Ⅰ frame masking effect, will lead to error propagation to the other frames of the video sequence, resulting in the decline in the quality of the entire video sequence. The video sequence of the single viewpoint Ⅰ frame is the first a two-dimensional image of the whole video sequence image groups; stereoscopic video sequence of Ⅰ frame refers to the pair of stereo images of the stereoscopic video sequence, containing the left and right view. In a typical video codec standards such as MPEG-4 and H.264, single view video sequence I frame airspace error concealment methods taken based on weighted pixel average median filter interpolation error concealment methods, the use of the weighted average of the lost block adjacent pixel to reconstruct the missing pixels. This method for a good masking effect of the smooth region of the video frame, but a region having a lot of detail and texture will result in the reconstructed image blur. Although there are many algorithms that take into account the object boundary pixels in the image, edge detection and direction discrimination technology, be able to recover the lost around the block a highly relevant boundary details, but most of these techniques using Sobel operator to detect more complex, the edge of the lines contains details powerless. The three-dimensional video processing technology is an important direction for future development of video technology, three-dimensional video I frame (containing the left and right view stereo image pairs) airspace error concealment technology has become a research focus of the stereoscopic video technology. Although some of the airspace of the three-dimensional video error concealment techniques can get a better reconstruction results, but most of the end of the block-based stereo matching disparity vector encoding decoding end decoded to reconstruct the damaged image . With the continuous improvement of the living standards, as well as the rapid development of high-definition display technology, the existing airspace error concealment methods to achieve the visual effects have been unable to meet the requirements of the people. More refined details of the video image optimization is still a quite challenging task, further research needs to be done. In this paper, two based on the details of the optimization of airspace error concealment algorithm. Error concealment algorithm by reading a lot of literature at home and abroad, and a lot of simulation, comprehensive complexity of the algorithm and experimental results and many other factors, the two based on the details of the optimization of airspace. This paper proposed and designed two pairs of video sequence error concealment algorithm based on the details of the optimization of airspace and simulation to achieve VC 6.0 platform. This work are: (1) Based on the airspace of the Hessian matrix of the half-pixel concealment algorithm is proposed for single viewpoint video sequence, in the case of single-view video sequence I frame damaged conventional intraframe airspace error concealment algorithm is proposed using the Hessian matrix of detecting edge lines of the image half-pixel interpolation, and then based on this, the further refinement of the image pixel information to obtain more accurate edge direction. Is pushed in the direction of the outer half-pixel interpolation of neighboring pixels, using a half-pixel interpolation point instead of the distance be conceal pixels farther available pixel points, to improve the mask accuracy and to optimize the image reconstruction effect. (2) of the stereoscopic video based on the pixel matching airspace error concealment algorithm is proposed for a stereoscopic video sequence. At the decoder, be damaged view conceal pixels around the pixel and another pixel in the view available for stereo matching. The proposed algorithm to the pixel matching primitives, the matching unit is accurate to one pixel, thus provide a basis for better details in the reconstructed image conditions; using the corrected image pairs epipolar constraints to achieve a more accurate stereo matching; using normalized grayscale cross-correlation function to match the criteria, set the threshold to reduce the shelter and the uneven distribution of light intensity, to further improve the accuracy of matching; match the settings window and search range, to achieve the purpose to obtain a more accurate visual difference. (3) the proposed algorithm to achieve the purpose of better recovery of the damaged image details. From the experimental results of the subjective and objective evaluation of the results of this paper, two based on the details of the optimization of airspace error concealment algorithm to alleviate the the reconstructed image details blur traditional methods and \or three-dimensional video sequence, the proposed algorithm can obtain better subjective visual effect and high PSNR value.

Related Dissertations
More Dissertations