电子稳像技术作为一种视频处理技术，具有广阔的应用前景。对于民用设备，不稳定的视频会造成观察者的视觉疲劳，而对于军用设备，不稳定的视频会产生系统的漏警和虚警等，所以视频质量的好坏不仅仅关系到人们生活，同时也关系到国家的安全。 本论文首先论述了基于TI的DM6467的电子稳像开发平台的搭建过程，U-boot的启动流程和Linux内核在DM6467上的移植过程，并实现网络文件系统（NFS）的挂载；其次，论述了电子稳像系统的基本原理，即采用基于快速鲁棒特征点（SURF）算法完成视频帧间图像的特征点匹配关系，利用空间-随机抽样一致性(SCRAMSAC)算法剔除误匹配特征点，利用Levenberg-Marguardt算法（LM算法）准确求出图像的运动估计矢量。通过参考帧更新策略对视频信息进行运动补偿，从而实现高精度的数字稳像系统；再次，介绍了该电子稳像的原理在DM6467上的实现，包括Codec Engine的生成、Server的配置及App的脚本文件的配置及程序的优化方法；此外，还绍了评价电子稳像结果的方法，包括主观评价和客观评价，客观评价又包括峰值信噪比和加权峰值信噪比两种方法；最后，给出本课题各阶段的实验结果并对该实验结果进行分析。 本课题中获得抖动的视频的方法是晃动摄像头，然后输入TMS320DM6467评估板，经过DM6467上稳像算法的处理，最后由LCD输出，实验结果表明，本课题在分辨率为720×576、帧率为25pfs的情况下能够稳定输出。

Electronic image stabilization technology as a video processing technology , has wide application prospects . For civilian equipment , unstable video will cause visual fatigue . And for military equipment, because of unstable video ， the system will produce leakage alarms and false alarms, etc., so the video quality is good or bad is not only related to people's lives, but also related to national security . In this paper , firstly discusses the process of building development platform which mounts electronic image stabilization based on TI's DM6467 , The process of booting U-boot and the process of transplanting Linux kernel on the DM6467, and to achieve the Network File System (NFS) is mounted. Secondly , discusses the basic principles of electronic image stabilization system, That is based on fast robust feature points (SURF) algorithm to complete the feature points of the inter relationship between frames of the video matching relations , use of space random sample consistency (SCRAMSAC) algorithm to eliminate mismatch feature points , and using the Levenberg-Marguardt (LM) algorithm obtained accurate motion vector . By updating the reference frame of video information policy to implement motion compensation and to achieve high-precision digital image stabilization system; Once again, introduced the principle of the electronic image stabilization achieved in the DM6467, including Codec Engine generation, configuration and optimization of procedures App Server configuration and script files; addition, we introduce methods of video quality assessment, including subjective evaluation and objective evaluation, objective evaluation also includes the peak signal to noise ratio (PSNR) and weighted peak signal to noise ratio(WPSNR); Finally, the experimental results for each stage of the subject and to analyze the results .This topic is obtained by shaking the camera shaky video, and then enter the TMS320DM6467 evaluation board, after the DM6467 image stabilization processing algorithms, the finally display in the LCD, the experimental results show that the subject in the resolution of 720 × 576, the frame rate 25pfs case to stabilize output .