视频信号压缩及图像稳定性算法的研究

发布时间：2018-06-28 02:12

本文选题：图像编码及压缩 + 图像目标识别及分割　；参考：《西安电子科技大学》2014年博士论文

【摘要】：随着视频技术的迅速发展及广泛应用，提供快速、有效及自动化的图像序列表达及处理方法已经成为了一个重要的研究领域，其中基于图像内容的方法，例如基于目标及特征区域的方法已经成为许多应用的首选。由于此类方法能够有效地消除图像序列中的“握手效应”，快速提取图像序列中的目标，以低比特率传输图像的形状、运动及纹理信息，并保持图像稳定性，因此它们在现代互联网、广播、电视、娱乐及第四代移动通讯等领域正被广泛采用。本论文介绍了作者对低比特率图像序列编码算法及框架的研究，重点研究了基于内容的图像目标分段、稳定性算法及实现框架设计。本文的目的是研究如何寻找简便高效的算法并提出一个集成这些算法的实现框架，使理论与实验成果走向实际应用。作者的主要研究成果包括：首先，在研究以往的目标分段算法的基础上，，提出了一种用于目标识别的自适应变化检测新算法。该算法采用一个三步法，能快速有效地把图像目标从背景中分离出来。第一步是依据亮度差及照度变化，分别把图像序列中的噪声和运动目标识别并分离出来；第二步是利用图像块、直方图及区域分类，把图像分割成为与运动目标相对应的区域；第三步是在前两步的基础上，进行形态边缘检测、轮廓分析及目标标识，以完成最终的目标识别、图像分段任务。其次，作者在上述图像目标识别、分割算法的基础上，设计了一个新的低比特率图像序列编码方案，该方案利用图像变化区域内的运动矢量信息、图像形状角点信息及无运动或准静止区域内的余留信息来完成高效视频压缩，其编解码性能优于传统的典型的编码算法。此外，本文针对实际应用中的图像序列不稳定现象（通常来自于摄像源），提出了一种新颖的图像运动补偿方法，该方法对来自于图像序列源的运动进行估计，并以补偿平移和旋转的方式来抵消此类运动。实验结果显示，该算法可以有效地稳定实时捕获的各类视频。为了验证本文提出和改进的有关算法，作者进行了大量的计算机模拟及实验，并同以往的传统经典方法进行了比较，说明了本文提出的若干方法取得了良好的效果。大量不同类型的实际图像序列实验也表明，本文提出的算法和方案性能可靠并优于文献中的典型算法，具有较好的应用前景。
[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.
【学位授予单位】：西安电子科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TN919.81

【参考文献】