基于层次化结构信息的视频重着色

发布时间：2019-04-09 06:23

【摘要】：视频重着色是一种通过重着色视频提高色盲用户对视频的可访问性的最直接有效的方法。它采用颜色转换函数改变视频帧的颜色,提高色盲用户获取颜色信息的能力,使被色盲用户混淆的颜色在重着色后能被感知到。近年来,随着多媒体搜索技术的快速发展,重着色可视化对象的研究越来越深入,成为帮助色盲用户感知可视化对象领域中的研究热点。本文的主要工作和创新点如下：1.概述了现有重着色算法中颜色转换函数的建立,并分析了现有算法的一些优点和不足之处。2.视频重着色是色盲重着色中一个具有挑战性的问题,目前的重着色算法在实现视频重着色过程中存在实时性问题和颜色不一致问题。视频重着色要对所有的图像帧进行重着色,要实现实时性就要求重着色视频帧的时间小于1/帧率,目前很难突破实时视频重着色问题。针对实时性问题,本文把基于CIELAB颜色空间的快速图像重着色算法应用到视频帧重着色中,通过两个简单的颜色旋转即可完成重着色,不仅重着色性能较好,而且需要的计算量比较小。快速图像重着色方法不能满足视频实时性的要求,本文提出一种颜色空间转换加速的方法缩短重着色时间以达到实时性要求,包括颜色空间量化和哈希表的使用。3.针对视频重着色存在的颜色不一致问题,提出一种视频层次化结构信息融合的方法,即“镜头+场景”的处理方式,目的是同时保持镜头和场景内的颜色一致性。由于镜头出现同一事物的可能性很大,在CIELAB颜色空间中保证镜头内所有帧的颜色转换函数相同。由于场景是由联系紧密的镜头组成,所以场景中不同镜头出现相同事物的可能性比较大,通过场景中镜头之间的颜色转换函数缓慢变化,实现整个场景的颜色一致性。通过本文方法可以实现实时视频重着色。
[Abstract]:Video re-coloring is one of the most direct and effective methods to improve the accessibility of color blind users to video by re-coloring video. The color conversion function is used to change the color of video frames so as to improve the ability of color blind users to obtain color information so that the color confused by color blind users can be perceived after re-coloring. In recent years, with the rapid development of multimedia search technology, the study of re-colored visual objects has become more and more in-depth, and has become a hot research topic in the field of color blind users' perception of visual objects. The main work and innovations of this paper are as follows: 1. This paper summarizes the establishment of the color conversion function in the existing re-coloring algorithms, and analyzes some advantages and disadvantages of the existing algorithms. 2. Video re-coloring is a challenging problem in blind color re-coloring. At present, there are real-time problems and color inconsistencies in the implementation of video re-coloring algorithms. In order to realize real-time performance, the time of re-coloring video frame is less than 1 / frame rate, so it is difficult to break through the problem of real-time video re-coloring. In order to solve the real-time problem, this paper applies the fast image re-coloring algorithm based on CIELAB color space to video frame re-coloring. Through two simple color rotation, the re-coloring can be completed, and not only the performance of re-coloring is better. Moreover, the amount of calculation required is relatively small. The fast image re-coloring method can not meet the real-time requirements of video. This paper proposes an accelerated color space conversion method to shorten the re-coloring time to meet the real-time requirements, including color space quantization and the use of hash tables. 3. In order to solve the problem of color inconsistency in video re-coloring, a method of video hierarchical information fusion is proposed, which is called "shot scene". The purpose of this method is to keep the color consistency between the shot and the scene at the same time. Since the lens is likely to have the same thing, the color conversion function of all frames in the lens is guaranteed to be the same in the CIELAB color space. Because the scene is composed of closely connected shots, it is more likely that different shots in the scene will appear the same thing. The color consistency of the whole scene can be achieved by slowly changing the color conversion function between the shots in the scene. Through this method, real-time video re-coloring can be realized.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【相似文献】