HEVC中面向繪制質(zhì)量的快速深度圖編碼研究
發(fā)布時間：2018-08-06 21:41
【摘要】：隨著三維顯示和交互多媒體系統(tǒng)的迅速發(fā)展，三維視頻系統(tǒng)將成為下一代視頻應(yīng)用的主要方向且日益受到人們關(guān)注。與傳統(tǒng)二維視頻相比，三維視頻系統(tǒng)帶來了良好的用戶視覺體驗，與此同時也使得視頻數(shù)據(jù)量成倍增加，因此急切地需要研究高性能的視頻壓縮算法。深度圖作為三維視頻系統(tǒng)中兩個視頻格式之一，研究高效的深度圖編碼算法是三維視頻系統(tǒng)中高性能視頻壓縮研究的重要組成部分，本文在分析深度圖編碼的研究現(xiàn)狀和存在的問題基礎(chǔ)上，著重研究HEVC中面向繪制質(zhì)量的深度圖像快速編碼算法，，主要內(nèi)容包括： 作為面向繪制質(zhì)量的深度圖像快速編碼研究的第一步，研究了深度圖像不同部分編碼失真對繪制虛擬視點失真的影響大小，得出具體結(jié)論：在相同的編碼失真情況下，深度值越大的點，產(chǎn)生的繪制失真越大；在紋理豐富區(qū)域點的失真對繪制的虛擬視點失真影響大，平坦區(qū)域點的失真對繪制的虛擬視點失真影響�。簧疃葓D像適當(dāng)?shù)钠交д婵梢蕴嵘L制虛擬視點圖像的質(zhì)量。這一結(jié)論是本文研究面向繪制質(zhì)量的高效深度圖像編碼的理論基礎(chǔ)。 針對HEVC編碼的CU（Coding Unit）結(jié)構(gòu)特征，提出了一種面向繪制質(zhì)量的深度圖像快速CU決策算法。首先，通過二次統(tǒng)計深度值劃分深度圖像LCU（Large Coding Unit）區(qū)域；然后，在深度圖像不同區(qū)域點的失真對繪制虛擬視失真影響大小的理論指導(dǎo)下，對已劃分區(qū)域的LCU進(jìn)行快速CU決策。實驗結(jié)果表明：該算法能夠有效地減少深度圖像在HEVC上的編碼時間，因為算法考慮了不同區(qū)域點的失真對繪制失真的影響，所以也保證了解碼后深度圖像的繪制質(zhì)量。 考慮到深度圖像特性以及HEVC幀內(nèi)編碼復(fù)雜度，在HEVC上提出了一種面向繪制質(zhì)量的深度圖像快速幀內(nèi)編碼算法。綜合考慮編碼預(yù)測準(zhǔn)確性和編碼計算復(fù)雜度，分三步來實現(xiàn)算法。首先，根據(jù)深度圖像特征對深度圖像LCU區(qū)域進(jìn)行預(yù)劃分；然后，在深度圖像不同區(qū)域點的編碼失真對繪制質(zhì)量影響大小的理論基礎(chǔ)上，進(jìn)行快速CU決策和幀內(nèi)PB（PredictBlock）選擇；最后，基于已劃分的PB區(qū)域進(jìn)行幀內(nèi)預(yù)測模式預(yù)先選擇。實驗結(jié)果表明：該方法在保證與HEVC原始幀內(nèi)編碼方法幾乎相同的繪制質(zhì)量的前提下，可平均提升深度圖編碼速度35%以上。
[Abstract]:With the rapid development of 3D display and interactive multimedia system, the 3D video system will become the main direction of the next generation video application and more and more people pay attention to it. Compared with the traditional two dimensional video, the 3D video system brings good user visual experience, and at the same time it makes the video data amount multiplied, so it is urgently needed. In order to study the high performance video compression algorithm, the depth map is one of the two video formats in the 3D video system. The study of the efficient depth map coding algorithm is an important part of the high performance video compression research in the 3D video system. Based on the analysis of the current research status and the existing problems of the depth map coding, this paper focuses on the research of HEVC The depth image fast encoding algorithm for rendering quality includes:
As the first step in the fast coding of deep image coding for drawing quality, the influence of the distortion of different parts of the depth image on the rendering of virtual view distortion is studied. The concrete conclusion is drawn: the higher the depth of the depth of the same coding distortion, the greater the distortion is, and the distortion in the rich region of the texture. The influence of the virtual view distortion is great, the distortion of the flat region point has little influence on the virtual view distortion, and the proper smoothness of the deep image can improve the quality of the virtual view image. This conclusion is the theoretical basis of the study of high efficiency deep image coding for drawing quality.
In view of the structure features of CU (Coding Unit) encoded by HEVC, a fast CU decision algorithm for rendering quality is proposed. Firstly, the depth image LCU (Large Coding Unit) region is divided by two statistical depth values, and then, under the theoretical guidance of the effect of the distortion of different region points in the depth image on the size of rendering virtual distortion. The experimental results show that the algorithm can effectively reduce the encoding time of the depth image on HEVC, because the algorithm takes into account the influence of the distortion of different regional points on the rendering distortion, so it also guarantees the rendering quality of the decoded depth image, the experimental results show that the algorithm can effectively reduce the encoding time of the depth image on HEVC.
Considering the characteristics of deep image and the complexity of HEVC intra coding, a fast intra coding algorithm for drawing quality is proposed on HEVC. The algorithm is divided into three steps, considering the accuracy of coding prediction and the complexity of coding calculation. First, the LCU region of the depth image is Predivided according to the feature of the depth image. Then, fast CU decision and intra PB (PredictBlock) selection are carried out on the basis of the coding distortion of different points in different regions of the depth image, and the intra frame prediction mode is pre selected based on the divided PB region. The experimental results show that the method is almost guaranteed to be encoded in the HEVC original frame. Under the same drawing quality, the depth map coding speed can be increased by more than 35%.
【學(xué)位授予單位】：上海大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TN919.81
本文編號：2169072
[Abstract]:With the rapid development of 3D display and interactive multimedia system, the 3D video system will become the main direction of the next generation video application and more and more people pay attention to it. Compared with the traditional two dimensional video, the 3D video system brings good user visual experience, and at the same time it makes the video data amount multiplied, so it is urgently needed. In order to study the high performance video compression algorithm, the depth map is one of the two video formats in the 3D video system. The study of the efficient depth map coding algorithm is an important part of the high performance video compression research in the 3D video system. Based on the analysis of the current research status and the existing problems of the depth map coding, this paper focuses on the research of HEVC The depth image fast encoding algorithm for rendering quality includes:
As the first step in the fast coding of deep image coding for drawing quality, the influence of the distortion of different parts of the depth image on the rendering of virtual view distortion is studied. The concrete conclusion is drawn: the higher the depth of the depth of the same coding distortion, the greater the distortion is, and the distortion in the rich region of the texture. The influence of the virtual view distortion is great, the distortion of the flat region point has little influence on the virtual view distortion, and the proper smoothness of the deep image can improve the quality of the virtual view image. This conclusion is the theoretical basis of the study of high efficiency deep image coding for drawing quality.
In view of the structure features of CU (Coding Unit) encoded by HEVC, a fast CU decision algorithm for rendering quality is proposed. Firstly, the depth image LCU (Large Coding Unit) region is divided by two statistical depth values, and then, under the theoretical guidance of the effect of the distortion of different region points in the depth image on the size of rendering virtual distortion. The experimental results show that the algorithm can effectively reduce the encoding time of the depth image on HEVC, because the algorithm takes into account the influence of the distortion of different regional points on the rendering distortion, so it also guarantees the rendering quality of the decoded depth image, the experimental results show that the algorithm can effectively reduce the encoding time of the depth image on HEVC.
Considering the characteristics of deep image and the complexity of HEVC intra coding, a fast intra coding algorithm for drawing quality is proposed on HEVC. The algorithm is divided into three steps, considering the accuracy of coding prediction and the complexity of coding calculation. First, the LCU region of the depth image is Predivided according to the feature of the depth image. Then, fast CU decision and intra PB (PredictBlock) selection are carried out on the basis of the coding distortion of different points in different regions of the depth image, and the intra frame prediction mode is pre selected based on the divided PB region. The experimental results show that the method is almost guaranteed to be encoded in the HEVC original frame. Under the same drawing quality, the depth map coding speed can be increased by more than 35%.
【學(xué)位授予單位】：上海大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2014
【分類號】：TN919.81
【參考文獻(xiàn)】
相關(guān)期刊論文 前1條
1 ;網(wǎng)聞[J];中國醫(yī)院院長;2013年10期
本文編號：2169072
本文鏈接：http://sikaile.net/kejilunwen/wltx/2169072.html
最近更新
教材專著
