本文選題：圖像編碼及壓縮 + 圖像目標(biāo)識(shí)別及分割��； 參考：《西安電子科技大學(xué)》2014年博士論文

【摘要】：隨著視頻技術(shù)的迅速發(fā)展及廣泛應(yīng)用，提供快速、有效及自動(dòng)化的圖像序列表達(dá)及處理方法已經(jīng)成為了一個(gè)重要的研究領(lǐng)域，其中基于圖像內(nèi)容的方法，例如基于目標(biāo)及特征區(qū)域的方法已經(jīng)成為許多應(yīng)用的首選。由于此類(lèi)方法能夠有效地消除圖像序列中的“握手效應(yīng)”，快速提取圖像序列中的目標(biāo)，以低比特率傳輸圖像的形狀、運(yùn)動(dòng)及紋理信息，并保持圖像穩(wěn)定性，因此它們?cè)诂F(xiàn)代互聯(lián)網(wǎng)、廣播、電視、娛樂(lè)及第四代移動(dòng)通訊等領(lǐng)域正被廣泛采用。 本論文介紹了作者對(duì)低比特率圖像序列編碼算法及框架的研究，重點(diǎn)研究了基于內(nèi)容的圖像目標(biāo)分段、穩(wěn)定性算法及實(shí)現(xiàn)框架設(shè)計(jì)。本文的目的是研究如何尋找簡(jiǎn)便高效的算法并提出一個(gè)集成這些算法的實(shí)現(xiàn)框架，使理論與實(shí)驗(yàn)成果走向?qū)嶋H應(yīng)用。 作者的主要研究成果包括：首先，在研究以往的目標(biāo)分段算法的基礎(chǔ)上，，提出了一種用于目標(biāo)識(shí)別的自適應(yīng)變化檢測(cè)新算法。該算法采用一個(gè)三步法，能快速有效地把圖像目標(biāo)從背景中分離出來(lái)。第一步是依據(jù)亮度差及照度變化，分別把圖像序列中的噪聲和運(yùn)動(dòng)目標(biāo)識(shí)別并分離出來(lái)；第二步是利用圖像塊、直方圖及區(qū)域分類(lèi)，把圖像分割成為與運(yùn)動(dòng)目標(biāo)相對(duì)應(yīng)的區(qū)域；第三步是在前兩步的基礎(chǔ)上，進(jìn)行形態(tài)邊緣檢測(cè)、輪廓分析及目標(biāo)標(biāo)識(shí)，以完成最終的目標(biāo)識(shí)別、圖像分段任務(wù)。 其次，作者在上述圖像目標(biāo)識(shí)別、分割算法的基礎(chǔ)上，設(shè)計(jì)了一個(gè)新的低比特率圖像序列編碼方案，該方案利用圖像變化區(qū)域內(nèi)的運(yùn)動(dòng)矢量信息、圖像形狀角點(diǎn)信息及無(wú)運(yùn)動(dòng)或準(zhǔn)靜止區(qū)域內(nèi)的余留信息來(lái)完成高效視頻壓縮，其編解碼性能優(yōu)于傳統(tǒng)的典型的編碼算法。 此外，本文針對(duì)實(shí)際應(yīng)用中的圖像序列不穩(wěn)定現(xiàn)象（通常來(lái)自于攝像源），提出了一種新穎的圖像運(yùn)動(dòng)補(bǔ)償方法，該方法對(duì)來(lái)自于圖像序列源的運(yùn)動(dòng)進(jìn)行估計(jì)，并以補(bǔ)償平移和旋轉(zhuǎn)的方式來(lái)抵消此類(lèi)運(yùn)動(dòng)。實(shí)驗(yàn)結(jié)果顯示，該算法可以有效地穩(wěn)定實(shí)時(shí)捕獲的各類(lèi)視頻。 為了驗(yàn)證本文提出和改進(jìn)的有關(guān)算法，作者進(jìn)行了大量的計(jì)算機(jī)模擬及實(shí)驗(yàn)，并同以往的傳統(tǒng)經(jīng)典方法進(jìn)行了比較，說(shuō)明了本文提出的若干方法取得了良好的效果。大量不同類(lèi)型的實(shí)際圖像序列實(shí)驗(yàn)也表明，本文提出的算法和方案性能可靠并優(yōu)于文獻(xiàn)中的典型算法，具有較好的應(yīng)用前景。
[Abstract]:With the rapid development and wide application of video technology, it has become an important research field to provide fast, effective and automated image sequence expression and processing. The method based on image content, such as the method based on target and feature area, has become the first choice for many applications. To eliminate the "handshake effect" in the image sequence, quickly extract the target in the image sequence, transmit the image shape, motion and texture information at low bit rate, and maintain the image stability, so they are being widely used in the fields of modern Internet, broadcasting, television, entertainment and the four generation mobile communication.
This paper introduces the author's research on the coding algorithm and framework of low bit rate image sequence, focusing on the content based image target segmentation, stability algorithm and implementation framework design. The purpose of this paper is to find a simple and efficient algorithm and put forward a framework to integrate these algorithms, so as to make the theoretical and experimental results. Go to practical application.
The main research results of the author include: first, on the basis of the study of the previous segmentation algorithm, a new adaptive change detection algorithm for target recognition is proposed. The algorithm uses a three step method to quickly and efficiently separate the image target from the background. The first step is based on the brightness difference and illumination variation, respectively. The noise and moving target in the image sequence are identified and separated. The second step is to divide the image into a region that corresponds to the moving target by using the image block, histogram and region, and the third step is to detect the shape edge, outline and identify the target on the basis of the first two steps, so as to complete the final target recognition. Image segmentation task.
Secondly, on the basis of the image target recognition and segmentation algorithm, a new low bit rate image sequence coding scheme is designed. The scheme uses the motion vector information in the image changing region, the image shape corner information and the residual information in the non motion or quasi stationary region to complete the efficient video compression, and its codec A typical coding algorithm that is superior to the traditional one.
In addition, a novel image motion compensation method is proposed for the image sequence instability (usually from the camera source) in practical applications. This method estimates the motion from the image sequence source and compensates for the translation and rotation of the motion. Experimental results show that the algorithm is effective. Stable and real-time capture of all kinds of video.
In order to verify the algorithm proposed and improved in this paper, a large number of computer simulations and experiments have been carried out, and compared with the traditional classical methods, it shows that some methods proposed in this paper have achieved good results. A large number of different types of actual image sequence experiments also show the performance of the proposed algorithm and scheme. It is reliable and superior to the typical algorithm in the literature, and has a good application prospect.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2014
【分類(lèi)號(hào)】：TN919.81

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 王相海;張洪為;李放;;遙感圖像高斯與椒鹽噪聲的PDE混合去噪模型研究[J];測(cè)繪學(xué)報(bào);2010年03期

2 曾德國(guó);熊輝;龍柯宇;唐斌;;基于相位差分的脈內(nèi)調(diào)制信號(hào)類(lèi)型識(shí)別[J];電子測(cè)量與儀器學(xué)報(bào);2009年10期

3 唐志峰;王詩(shī)俊;楊樹(shù)元;;一種高精度的壓縮域視頻目標(biāo)分割算法[J];電子與信息學(xué)報(bào);2007年12期

4 楊任爾;陳懇;何加銘;;基于預(yù)測(cè)的多描述圖像編碼冗余插入的研究[J];光電工程;2007年10期

5 王智慧;王敬東;李鵬;張春;;一種基于KLT-RANSAC全局運(yùn)動(dòng)估計(jì)的電子穩(wěn)像算法研究[J];光電子技術(shù);2012年01期

6 王小平;靜大海;;基于二維局域波和角點(diǎn)匹配的多模態(tài)圖像配準(zhǔn)[J];電子設(shè)計(jì)工程;2013年04期

7 張慧芳;金文光;;低碼率下CBC算法中位平面編碼的新方法[J];浙江大學(xué)學(xué)報(bào)(理學(xué)版);2009年01期

8 楚瀛;田淞;張桂林;張熠;;基于圖像邊緣特征的前景背景分割方法[J];華中科技大學(xué)學(xué)報(bào)(自然科學(xué)版);2008年05期

9 李業(yè)偉;華臻;李晉江;;采用顯著區(qū)域匹配的圖像拼接算法[J];計(jì)算機(jī)工程與應(yīng)用;2012年10期

10 武艷美;肖陽(yáng)輝;;基于特征點(diǎn)匹配的全局運(yùn)動(dòng)估計(jì)[J];計(jì)算機(jī)工程;2011年22期

本文編號(hào)：2076261