基于块匹配的电子稳像技术研究

发布时间：2019-05-09 05:16

【摘要】：随着科学技术的进步,图像作为一种可视化、直接的信息在运动载体上的需求以及应用变得越来越多,但是在实际操作过程中,运动载体的拍摄装置在运动过程中会出现由于载体自身机械振动或者自然环境等因素引起图像抖动、图像画面不稳定的现象,因此稳像技术应运而生,其中对电子稳像技术的研究越来越多。电子稳像是指用数字图像处理的方法实现对图像抖动的估计和补偿,从而使得图像保持稳定。本文对电子稳像技术进行研究,以求兼顾稳像的精度和稳像的快速性。本文首先根据相机的成像模型,推导了相机运动情况下空间坐标和图像坐标之间的关系,从而得到图像帧间运动模型。对电子稳像的基本原理进行了描述,介绍了电子稳像的三个环节:运动估计、运动滤波、运动补偿的基本任务,给出了电子稳像质量的评价方法。然后分析研究了常见的运动估计算法:块匹配法、灰度投影法、位平面法、特征匹配法、光流法,常见的运动滤波算法:均值滤波法、卡尔曼滤波法、曲线拟合法,其中块匹配的特点可以较完美地契合本课题的目的,然后对块匹配法的搜索策略进行详细研究,包括全搜索法、三步搜索法、四步搜索法、菱形搜索法和六边形搜索法。最后,本文提出基于块匹配的电子稳像系统,选用性能较优的六边形搜索,并实现改进,此外,由于卡尔曼滤波只需要前一时刻的信息,在实时性要求高的系统中应用非常多,因此本文选用卡尔曼滤波进行滤波实现。在MATLAB环境下实现以上流程的稳像算法对视频图像进行了稳像处理,实验结果表明该算法具有很好的稳像效果。
[Abstract]:With the progress of science and technology, image as a kind of visualization, the demand and application of direct information on the moving carrier become more and more, but in the actual operation process, In the process of motion, the shooting device of the moving carrier will cause the image jitter due to the mechanical vibration of the carrier itself or the natural environment, and the image picture will be unstable, so the image stabilization technology emerges as the times require. Among them, there is more and more research on electronic image stabilization technology. Electronic image stabilization refers to the estimation and compensation of image jitter by digital image processing, so that the image is stable. In this paper, the electronic image stabilization technology is studied in order to take into account the accuracy of image stabilization and the rapidity of image stabilization. In this paper, according to the imaging model of the camera, the relationship between the spatial coordinates and the image coordinates under the condition of camera motion is derived, and the inter-frame motion model of the image is obtained. In this paper, the basic principle of electronic image stabilization is described, and the basic tasks of motion estimation, motion filtering and motion compensation are introduced. The evaluation method of the quality of electronic image stabilization is given. Then the common motion estimation algorithms are analyzed and studied: block matching, gray projection, bit plane, feature matching, optical flow, and common motion filtering algorithms: mean filtering, Kalman filtering, curve fitting, and so on. The characteristics of block matching can fit the purpose of this topic perfectly. Then the search strategy of block matching method is studied in detail, including full search method, three-step search method, four-step search method, diamond search method and hexagon search method. Finally, an electronic image stabilization system based on block matching is proposed in this paper. The hexagonal search with better performance is selected and improved. In addition, because Kalman filter only needs the information of the previous moment, it is widely used in the system with high real-time requirements. Therefore, Kalman filter is selected to realize the filtering in this paper. The image stabilization algorithm of the above flow is implemented in MATLAB environment to stabilize the video image. The experimental results show that the algorithm has a good image stabilization effect.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】