基于自适应补偿的电子稳像方法.docVIP

基于自适应补偿的电子稳像方法 尹丽华李范鸣刘士建 中国科学院上海技术物理研宄所中国科学院大学 上海科技大学中国科学院红外探测与成像技术重点 实验室 基于车载的光电系统在采集图像的过程中，载车的随机振动会导致视频序列发 牛.抖动，需要通过稳像技术来消除。为解决传统电子稳像方法中存在帧间跳变、 机械性选择参考帧，以及边缘信息丢失的问题，提出了一种基于自适疲补偿的 电子稳像方法。首先，白适应地更新参考帧，后续帧与更新的参考帧进行特征点 的提取和匹配，并通过参数迭代方法求解相对于第1帧的全局运动参数;然后， 区分不同的全局运动参数:旋转角度、水平、垂直方向的偏移量，并通过运动滤 波从全局运动参数中分离出扫描运动，补偿随机抖动;最后，采用后向映射方式 实现运动补偿，并运用图像镶嵌，即利用参考基准帧的信息对校正后图像的边 缘信息进行补偿。实验结果表明:稳像后视频序列有很好的视觉稳定效果，且提 高了峰值信噪比(PSNR)和帧间变换保真度(ITF)。本文的方法具有很好的鲁 棒性，在不同的视频场景下都可以实现稳像，可以满足稳像的精度要求。 关键词： 电子稳像;白适应;参数迭代;图像镶嵌;PSNR; 基金：国家863计划资助项B (No. 2011AA7031002G) Electronic image stabilization method based on adaptive compensation Shanghai Institute of Technical Physics of The Chinese Academy of Sciences; Abstract： Tn order to solve the problems of interframe hopping, mechanically selecting reference frame, missing edge information in traditional electronic image stabilization method, an electronic image stabilization method based on adaptive compensation is proposed. Firstly, the reference frame is updated adaptively, and feature points are extracted and matched for the follow-up frames and updated reference frame, then the global motion parameters relative to the first frame can be solved through parameter iteration method. Secondly, different global motion parameters are distinguished, such as rotation angle, horizontal and vertical offset, the scanning movement was isolated from the global motion parameters by filtering and the random jitter can be compensated. Finally, motion compensation can be achieved by backward mapping and image mosaic, namely compensating the image edge information according to the basis reference frame of information. Experimental results indicate that the video sequence after image stabilization has good stable visual effect with higher PSNR and ITF. The method has good robustness, and can achieve image stabilization under different video scenes, which meets the precision requirements. Keyword： electronic image stabilization; adaptive; parameter iteration; image mosaic; P