基于冗余優(yōu)化的多描述三維圖像編碼
[Abstract]:3D technology is widely used in many fields, such as industry, education, spaceflight, medical treatment and so on. It is also the direction of multimedia development in the future. However, stereo technology requires the number of views, followed by the problem of data volume. Therefore, efficient image compression has become a hot research to counter the surge of data. In this paper, a combination of texture image and depth image is used to represent three-dimensional image. The emergence of multi-description coding solves the problem of serious deterioration of quality caused by network delay and packet loss in traditional channels. Based on the method of multi-description, this paper applies it to the coding and decoding of 3D images. The main works are as follows: (1) in 3D image technology, the depth image represents the distance between the object and the camera in the scene. Combined with the existing multi-view images, the images of any virtual location can be synthesized effectively, and the information of the whole scene can be displayed more completely. To some extent, some high-frequency components of depth image have little effect on the quality of composite image, especially for the multi-description coding method, the information will also cause a waste of bit rate for compression and transmission. This paper presents a method of multi-description coding based on redundancy removal of depth images in DCT domain. Based on the characteristics of DCT coefficients, a Lagrangian optimization algorithm is proposed to determine how much redundant information is removed in DCT domain. At the decoding end, an adaptive zero-fill scheme is proposed to reconstruct depth images. After coding and decoding and optimizing algorithms, the rate distortion performance of multi-description coding methods is improved effectively. (2) traditional multi-description coding methods often do not fully consider the characteristics of human visual system. The perception of image quality depends largely on the human eye. In this paper, a multi-description depth image coding method based on human visual system is proposed based on minimum perceptual difference (JND) information. The JND model of depth image is introduced into the multi-description coding of depth image, and the redundant information is optimized. Considering the characteristics of human visual system, the redundant information of depth image is readjusted, and some distortion of human eye sensitivity is paid attention to. After coding and optimization, better reconstruction quality can be obtained at the same bit rate. (3) in order to apply the proposed scheme to 3D image coding and decoding, In this paper, a multi-view JND model is constructed by combining the depth images and texture images of the two angles of view, and the quality evaluation of the synthetic images shows that the MJND model can accommodate more distortion. At the same time, combining the characteristics of texture image and the proposed MJND model, a coding scheme for texture image is proposed, which makes it more anti-jamming in the transmission process. The depth image is encoded and decoded according to the multi-description scheme based on human visual system. By evaluating the quality of synthetic image, better visual effect and reconstruction quality can be obtained in both subjective and objective aspects.
【學(xué)位授予單位】:北京交通大學(xué)
【學(xué)位級(jí)別】:碩士
【學(xué)位授予年份】:2017
【分類號(hào)】:TN919.81
【相似文獻(xiàn)】
相關(guān)期刊論文 前10條
1 張世輝;張煜婕;孔令富;;一種基于深度圖像的自遮擋檢測(cè)方法[J];小型微型計(jì)算機(jī)系統(tǒng);2010年05期
2 趙翠蓮;施曉磊;榮堅(jiān);范志堅(jiān);;基于坐標(biāo)軸投影勘察的深度圖像樹型分割[J];光學(xué)精密工程;2011年06期
3 許小艷;廖曉峰;劉然;易琳;張小云;;基于深度圖像繪制的視圖合成[J];系統(tǒng)仿真學(xué)報(bào);2011年10期
4 江文婷;龔小謹(jǐn);劉濟(jì)林;;顏色指導(dǎo)的深度圖像升采樣算法的對(duì)比性研究[J];杭州電子科技大學(xué)學(xué)報(bào);2014年01期
5 張濤,平西建,柳葆芳,邵美珍;一種深度圖像中的表面曲率估計(jì)算法[J];數(shù)據(jù)采集與處理;2001年01期
6 謝凱,湯曉安,郝建新,蔡宣平;一種基于多幅深度圖像的混合建模方法[J];小型微型計(jì)算機(jī)系統(tǒng);2003年01期
7 婁嘉鵬,李德華,吳險(xiǎn)峰,薛雷;深度圖像的平滑[J];計(jì)算機(jī)與數(shù)字工程;2003年01期
8 代明睿;張曉鵬;李紅軍;;基于高噪聲深度圖像的樹木模型重建[J];中國體視學(xué)與圖像分析;2010年02期
9 杜霄鵬;郝建平;李星新;楊俊;;基于單一深度圖像的人體姿態(tài)實(shí)時(shí)識(shí)別技術(shù)研究[J];計(jì)算機(jī)與現(xiàn)代化;2012年04期
10 羅莎莎;郭太良;;一種二維圖像序列的深度圖像生成方法[J];計(jì)算機(jī)與現(xiàn)代化;2012年05期
相關(guān)會(huì)議論文 前2條
1 孫曉蘭;趙惠潔;祝世平;;基于深度圖像的邊緣檢測(cè)方法研究[A];2004全國光學(xué)與光電子學(xué)學(xué)術(shù)研討會(huì)、2005全國光學(xué)與光電子學(xué)學(xué)術(shù)研討會(huì)、廣西光學(xué)學(xué)會(huì)成立20周年年會(huì)論文集[C];2005年
2 徐秀英;劉卓;王博亮;;一種基于三角函數(shù)的深度圖像曲面擬合方法[A];第十二屆全國圖象圖形學(xué)學(xué)術(shù)會(huì)議論文集[C];2005年
相關(guān)重要報(bào)紙文章 前1條
1 深圳商報(bào)記者 徐明天 實(shí)習(xí)生 桂馨;電視機(jī)可借助電力線上網(wǎng)[N];深圳商報(bào);2010年
相關(guān)博士學(xué)位論文 前8條
1 苗丹;三維視頻壓縮、傳輸、渲染相關(guān)技術(shù)的研究[D];中國科學(xué)技術(shù)大學(xué);2015年
2 譚筠;結(jié)構(gòu)化環(huán)境下基于信息融合的道路場(chǎng)景感知技術(shù)研究[D];國防科學(xué)技術(shù)大學(xué);2014年
3 范濤;基于壓縮感知深度圖像處理及三維編碼系統(tǒng)優(yōu)化方法[D];上海大學(xué);2017年
4 李世飛;基于深度圖像的三維目標(biāo)識(shí)別技術(shù)研究[D];國防科學(xué)技術(shù)大學(xué);2010年
5 宋希彬;基于RGB-D信息的深度圖像增強(qiáng)關(guān)鍵技術(shù)研究[D];山東大學(xué);2017年
6 丁雅斌;隱式表達(dá)三維模型流水線的關(guān)鍵技術(shù)研究[D];天津大學(xué);2007年
7 許小艷;多視點(diǎn)視頻中的視圖合成[D];重慶大學(xué);2011年
8 丁益洪;立體足跡三維曲面分割與識(shí)別[D];解放軍信息工程大學(xué);2005年
相關(guān)碩士學(xué)位論文 前10條
1 楊艷;基于深度圖像的三維重建方法研究[D];北京建筑大學(xué);2015年
2 王國強(qiáng);基于深度圖像的人體部位識(shí)別及動(dòng)作識(shí)別[D];華南理工大學(xué);2015年
3 龐云沖;基于機(jī)器學(xué)習(xí)思想的深度圖像遮擋邊界檢測(cè)方法研究[D];燕山大學(xué);2015年
4 張博文;基于深度圖的駕駛員頭部姿態(tài)分析[D];大連海事大學(xué);2015年
5 張?jiān)汽?基于深度圖像的玉米品種識(shí)別研究[D];河北農(nóng)業(yè)大學(xué);2015年
6 斯曉華;高效低復(fù)雜度三維視頻編碼技術(shù)研究[D];上海大學(xué);2015年
7 楊炳偉;基于深度圖像的三維重建研究[D];北京建筑大學(xué);2015年
8 崔穎;基于Kinect的人群識(shí)別與跟蹤系統(tǒng)[D];復(fù)旦大學(xué);2013年
9 羅浩;RGB-D視覺內(nèi)容理解及其應(yīng)用[D];華北電力大學(xué);2015年
10 王潤(rùn)元;基于深度圖像和彩色圖像融合的頭姿估計(jì)算法研究[D];北京工業(yè)大學(xué);2015年
,本文編號(hào):2405730
本文鏈接:http://www.sikaile.net/kejilunwen/xinxigongchenglunwen/2405730.html