【摘要】：隨著互聯(lián)網(wǎng)信息應(yīng)用領(lǐng)域的拓寬,圖像處理技術(shù)已成為計算機(jī)視覺領(lǐng)域內(nèi)一個重要的應(yīng)用技術(shù)。為了減少圖像的存儲空間和傳輸過程中所需要的時間,需要消除像素之間的數(shù)據(jù)冗余對圖像進(jìn)行壓縮。因此,圖像壓縮是數(shù)字圖像處理中一個重要研究方向。JPEG已經(jīng)成為廣泛應(yīng)用于靜止圖像的壓縮標(biāo)準(zhǔn),是一種變換編碼方法。JPEG最大的優(yōu)勢是可以根據(jù)用戶的需求自由地選擇壓縮比,在圖像質(zhì)量與代價之間作出決擇。但因壓縮過程中量化取整操作造成數(shù)據(jù)丟失,使解壓圖像產(chǎn)生塊效應(yīng),出現(xiàn)明顯的壓縮痕跡。MPEG壓縮標(biāo)準(zhǔn)是一種針對視頻的壓縮方法,其主要運用于視頻數(shù)據(jù)的存儲,廣播電視和視頻流的網(wǎng)絡(luò)傳輸。MPEG采用類似于靜止圖像的JPEG壓縮方法去除幀內(nèi)冗余,能夠獲得高壓縮比,達(dá)到減少存儲空間和傳輸時間的目的。但高壓縮比也意味著數(shù)據(jù)的丟失,導(dǎo)致解碼視頻質(zhì)量較差,具有明顯的壓縮痕跡。本文主要利用總變分?jǐn)?shù)學(xué)模型,以靜止圖像與視頻為研究對象,對JPEG和MPEG壓縮圖像進(jìn)行解碼,達(dá)到減輕解碼圖像的壓縮痕跡的目的,發(fā)展最優(yōu)化數(shù)值算法。本文所做的工作有:(1)簡單介紹JPEG和MPEG解碼圖像出現(xiàn)壓縮痕跡的原因,概述了國內(nèi)外關(guān)于消除圖像壓縮量化噪聲的研究現(xiàn)狀。(2)介紹了標(biāo)準(zhǔn)JPEG編碼算法,為方便建模,構(gòu)造JPEG編碼的數(shù)學(xué)模型,根據(jù)DCT系數(shù)的量化區(qū)間得到原圖像的一個先驗條件。利用二維總變分方法的基本理論,給出基于總變分消除量化噪聲的最優(yōu)化模型。最后利用原-對偶方法求解優(yōu)化模型,得到后處理的總變分JPEG解碼圖像。(3)介紹了標(biāo)準(zhǔn)MPEG視頻編碼算法,同樣根據(jù)量化取整得到一個量化區(qū)間。根據(jù)三維視頻總變分方法的基本理論,分別構(gòu)造耦合時間維度的總變分模型與分離時間維度的總變分模型,將兩種總變分模型轉(zhuǎn)化為最小最大問題,再采用原-對偶方法求解兩種模型,得到兩種后處理MPEG解碼視頻。
[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.
【學(xué)位授予單位】：昆明理工大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP391.41

【參考文獻(xiàn)】

相關(guān)期刊論文 前3條

1 曹文倫;史忠科;封建湖;;基于多項式擬合的交通圖像編碼方法[J];計算機(jī)工程與應(yīng)用;2006年19期

2 王相海,張福炎;靜態(tài)圖像編碼研究進(jìn)展[J];計算機(jī)研究與發(fā)展;2001年11期

3 李海波;模型基圖像編碼[J];通信學(xué)報;1993年02期

，

本文編號：2381092