论文部分内容阅读
以室外视频监控中帧间平移运动为研究对象,提出一种基于平衡位置修正与自适应存储帧更新的实时全分辨率数字稳像方法。首先利用卡尔曼滤波在线估计镜头的主动运动,预先确定抖动图像补偿后的平衡位置;然后针对图像补偿后产生的空白区域,通过选择合适的参考帧及预存帧构成存储帧队列,根据存储帧队列的覆盖范围对平衡位置进行修正,实现修正后空白区域像素的全部拼接填补;最后以修正后的平衡位置为基准,根据同时满足左上、右上、左下和右下4个偏移方向的原则对参考帧队列进行自适应在线更新,用最新采集帧替换时间上较早、空间内容上冗余的参考帧,使得存储帧队列在有限长度内保证尽可能大的空间覆盖。本文提出的方法可以满足实时数字全分辨率稳像的要求,适用于集成式、嵌入式等各类视频监控系统。
Taking the pan-tilt movement in outdoor video surveillance as the research object, a real-time full-resolution digital image stabilization method based on the balance position correction and adaptive storage frame updating is proposed. Firstly, Kalman filter is used to estimate the active motion of the lens online and the equilibrium position of the image after compensation is determined in advance. Then, for the blank area generated by the image compensation, the storage frame queue is constituted by choosing the appropriate reference frame and pre-stored frame, Of the coverage area to correct the position of the balance to achieve the corrected all the pixels in the blank area stitching fill; Finally, based on the corrected equilibrium position, according to the same time to meet the upper left, upper right, lower left and right under the 4 offset direction of the principle of reference Frame queue for adaptive online update, with the latest acquisition of the frame to replace the earlier in time, spatial content redundant reference frame, making the storage frame queue to ensure a limited length of the largest possible space coverage. The proposed method can meet the requirements of real-time digital full-resolution image stabilization and is suitable for all kinds of video surveillance systems such as integrated and embedded systems.