基于总变分的图像解码方法研究
发布时间：2018-12-15 18:27
【摘要】：随着互联网信息应用领域的拓宽,图像处理技术已成为计算机视觉领域内一个重要的应用技术。为了减少图像的存储空间和传输过程中所需要的时间,需要消除像素之间的数据冗余对图像进行压缩。因此,图像压缩是数字图像处理中一个重要研究方向。JPEG已经成为广泛应用于静止图像的压缩标准,是一种变换编码方法。JPEG最大的优势是可以根据用户的需求自由地选择压缩比,在图像质量与代价之间作出决择。但因压缩过程中量化取整操作造成数据丢失,使解压图像产生块效应,出现明显的压缩痕迹。MPEG压缩标准是一种针对视频的压缩方法,其主要运用于视频数据的存储,广播电视和视频流的网络传输。MPEG采用类似于静止图像的JPEG压缩方法去除帧内冗余,能够获得高压缩比,达到减少存储空间和传输时间的目的。但高压缩比也意味着数据的丢失,导致解码视频质量较差,具有明显的压缩痕迹。本文主要利用总变分数学模型,以静止图像与视频为研究对象,对JPEG和MPEG压缩图像进行解码,达到减轻解码图像的压缩痕迹的目的,发展最优化数值算法。本文所做的工作有:(1)简单介绍JPEG和MPEG解码图像出现压缩痕迹的原因,概述了国内外关于消除图像压缩量化噪声的研究现状。(2)介绍了标准JPEG编码算法,为方便建模,构造JPEG编码的数学模型,根据DCT系数的量化区间得到原图像的一个先验条件。利用二维总变分方法的基本理论,给出基于总变分消除量化噪声的最优化模型。最后利用原-对偶方法求解优化模型,得到后处理的总变分JPEG解码图像。(3)介绍了标准MPEG视频编码算法,同样根据量化取整得到一个量化区间。根据三维视频总变分方法的基本理论,分别构造耦合时间维度的总变分模型与分离时间维度的总变分模型,将两种总变分模型转化为最小最大问题,再采用原-对偶方法求解两种模型,得到两种后处理MPEG解码视频。
[Abstract]:With the widening of the application field of Internet information, image processing technology has become an important application technology in the field of computer vision. In order to reduce the storage space of the image and the time needed in the transmission process, it is necessary to eliminate the data redundancy between pixels to compress the image. Therefore, image compression is an important research direction in digital image processing. JPEG has become a widely used still image compression standard. The biggest advantage of JPEG is that it can freely choose the compression ratio according to the needs of the user and make a choice between the image quality and the cost. However, due to the loss of data caused by the quantization rounding operation in the compression process, the decompression image has block effect and obvious compression trace. MPEG compression standard is a compression method for video, which is mainly used in the storage of video data. MPEG uses JPEG compression method similar to still image to remove intra-frame redundancy, which can achieve high compression ratio and reduce storage space and transmission time. But high compression ratio also means the loss of data, resulting in poor quality of decoded video, with obvious compression trace. In this paper, the general variational fractional model is used to decode the JPEG and MPEG compressed images, taking the still image and video as the research object, so as to reduce the compression trace of the decoded image and to develop the optimal numerical algorithm. The main works of this paper are as follows: (1) the reasons for the compression traces of JPEG and MPEG decoded images are briefly introduced, and the research status of eliminating image compression quantization noise at home and abroad is summarized. (2) the standard JPEG coding algorithm is introduced, which is convenient for modeling. The mathematical model of JPEG coding is constructed and a priori condition of the original image is obtained according to the quantization interval of the DCT coefficient. Based on the basic theory of 2-D total variational method, an optimization model for eliminating quantization noise based on total variation is presented. Finally, the optimization model is solved by primal-duality method, and the post-processed total variational JPEG decoded image is obtained. (3) Standard MPEG video coding algorithm is introduced, and a quantization interval is also obtained according to quantization rounding. According to the basic theory of 3D video total variation method, the total variational model of coupling time dimension and the total variational model of separating time dimension are constructed, respectively, and the two general variational models are transformed into the minimum maximum problem. Then two kinds of post-processing MPEG decoded video are obtained by using primal-dual method to solve the two models.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
本文编号：2381092
[Abstract]:With the widening of the application field of Internet information, image processing technology has become an important application technology in the field of computer vision. In order to reduce the storage space of the image and the time needed in the transmission process, it is necessary to eliminate the data redundancy between pixels to compress the image. Therefore, image compression is an important research direction in digital image processing. JPEG has become a widely used still image compression standard. The biggest advantage of JPEG is that it can freely choose the compression ratio according to the needs of the user and make a choice between the image quality and the cost. However, due to the loss of data caused by the quantization rounding operation in the compression process, the decompression image has block effect and obvious compression trace. MPEG compression standard is a compression method for video, which is mainly used in the storage of video data. MPEG uses JPEG compression method similar to still image to remove intra-frame redundancy, which can achieve high compression ratio and reduce storage space and transmission time. But high compression ratio also means the loss of data, resulting in poor quality of decoded video, with obvious compression trace. In this paper, the general variational fractional model is used to decode the JPEG and MPEG compressed images, taking the still image and video as the research object, so as to reduce the compression trace of the decoded image and to develop the optimal numerical algorithm. The main works of this paper are as follows: (1) the reasons for the compression traces of JPEG and MPEG decoded images are briefly introduced, and the research status of eliminating image compression quantization noise at home and abroad is summarized. (2) the standard JPEG coding algorithm is introduced, which is convenient for modeling. The mathematical model of JPEG coding is constructed and a priori condition of the original image is obtained according to the quantization interval of the DCT coefficient. Based on the basic theory of 2-D total variational method, an optimization model for eliminating quantization noise based on total variation is presented. Finally, the optimization model is solved by primal-duality method, and the post-processed total variational JPEG decoded image is obtained. (3) Standard MPEG video coding algorithm is introduced, and a quantization interval is also obtained according to quantization rounding. According to the basic theory of 3D video total variation method, the total variational model of coupling time dimension and the total variational model of separating time dimension are constructed, respectively, and the two general variational models are transformed into the minimum maximum problem. Then two kinds of post-processing MPEG decoded video are obtained by using primal-dual method to solve the two models.
【学位授予单位】：昆明理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41
【参考文献】
相关期刊论文 前3条
1 曹文伦;史忠科;封建湖;;基于多项式拟合的交通图像编码方法[J];计算机工程与应用;2006年19期
2 王相海,张福炎;静态图像编码研究进展[J];计算机研究与发展;2001年11期
3 李海波;模型基图像编码[J];通信学报;1993年02期
，本文编号：2381092
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2381092.html
