基于下采样灰度投影的电子稳像算法研究

摘要

为了快速、准确地稳定矿井车载摄像系统获取的抖动图像,提出了一种基于下采样灰度投影的电子稳像算法.首先对图像进行直方图均衡化处理,以提高图像对比度;为减少算法运算量,削减局部运动物体的影响,通过下采样方法对图像进行多分辨率分层,选取相邻2帧低分辨率图像及其预定子区域分别进行灰度投影运算,并对运算结果进行加权求和处理,获得最终的全局运动矢量;最后根据卡尔曼滤波获取的补偿量,通过自适应相邻帧补偿方法对图像进行运动补偿,以解决误差累积问题.实验结果表明,所提算法在准确度和运算时间上都优于传统灰度投影算法,且能实现长时间连续稳像.%In order to stabilize dithering images obtained by mine vehicle-mounted recording system fast and accurately,an electronic image stabilization algorithm based on subsample gray-scale projection was proposed.Firstly,the image is processed by histogram equalization to improve image contrast.In order to reduce computation amount of the algorithm and weaken impact of local moving objects,the image is multi-resolution layered by the method of subsample,and the adjacent two frames of the low resolution image and its predetermined sub-regions are selected for gray-scale projection operation.The global motion vector is obtained by weighted summing process of the operation results.Finally,the image is compensated by adaptive adjacent frame compensation method according to compensation amount obtained by Kalman filtering.The experimental results show that the proposed algorithm is superior to traditional gray-scale projection algorithm in terms of accuracy and computation time,and can achieve long time continuous image stabilization.