由形态学边缘模式匹配实现数字稳像

针对在一些实际应用场合摄像机设备的震动造成的视频序列失稳,提出了一种基于形态学边缘模式匹配的数字稳像方法.该方法首先利用形态学方法提取视频图像边缘,然后利用当前帧子块与参考帧子块的边缘进行特征匹配,以确定当前帧子块相对于参考帧的局部运动矢量.对得到的局部运动矢量进行分析判决,以确定该局部运动矢量是否为真正的抖动偏移量,消除视频帧本身因存在局部运动目标、低对比度、纹理区域等因素对全局运动矢量估计的影响;最后,通过判别确认的局部运动矢量确定全局运动矢量并进行运动补偿,实现数字稳像.仿真实验表明,由于该方法的块局部运动矢量采用1 bit边缘进行最大相关值估计,因而复杂度低,而采用异常情况判决准则消除了其对真正运动矢量估计的影响,稳像精度高,可应用于因摄像头平移震动等因素造成的视频序列失稳.

Stabilization of digital image by morphological edge pattern matching

To eliminate the image jittering from the vibration caused by the translation motion of a camera, a method of digital image stabilization based on morphological edge pattern matching is proposed. In this method, morphological edge matching is used to estimate the local movement vectors based on the 1 bit edge maximal correlation. Since correlation calculation is logical operator instead of pixel-wise gray arithmetic operators, the complexity of proposed method is greatly reduced. Moreover, in order to remov...