一种基于改进光流法的电子稳像算法

针对煤矿井下机载抖动视频序列低照度、低清晰度以及易受噪声影响的特点，提出了一种基于改进光流法的电子稳像算法。算法利用SUSAN算子获取帧间匹配特征点，并提出基于金字塔改进光流模型，实现抖动视频序列的全局运动矢量计算。为克服机载图像中局部运动目标和移动背景的影响，利用RANSAC算法去除误匹配点，提出稳像参考帧的动态更新机制。针对矿下机载摄像抖动的不确定性，运用均值滤波求解运动补偿矢量，在保留主动运动的同时过滤随机抖动。理论分析和实验结果表明，提出的算法不仅保留了光流法获取运动矢量的精度，而且提高了算法的实时性，对于煤矿井下机载抖动视频序列具有较好的稳像效果。

An electronic image stabilization algorithm based on improved optical flow method

To improve the quality of images captured by moving underground cameras with the characteristics of low resolution，low illumination and being easily affected by noise，a new electronic image stabilization algorithm based on improved optical flow was proposed.Firstly，by extracting inter-frame matching feature points with the SUSAN algorithm，an improved optical flow method based on pyramid was proposed to calculate the global motion vectors.Furthermore，in order to overcome the influence of moving objects and moving background within images，the RANSAC algorithm was used to eliminate the error matching points，and a dynamic update mechanism for reference frame was proposed.Considering the feature of uncertain jitter from underground moving carriers，the vectors of motion compensation were solved by the mean filter algorithm，which can retain the active movement while filtering the random jitter.Theoretical analysis and experimental results show that the proposed method can not only retain the processing precision for motion vector by pyramid-based optical flow method，but also improve the real-time property of the algorithm.It has a better effect on stabilizing the coal mine underground jitter video sequence.