摘要
针对灰度投影法在灰度值单一、对比度差的图像序列中稳像精确度不足,改进基于灰度投影的运动估计算法,先采用卡尔曼滤波将采样图像进行预处理以去除噪声,再将每帧图像分成4个子区域,按原方法对各子区域内像素点进行灰度投影,计算出子区域在x轴和y轴上的偏移量,分别平均得到整体偏移量,用插值放大的方法进行运动补偿。选取日本九州大学图片数据库中图像序列的相邻两帧进行仿真试验,结果表明,改进算法的峰值信噪比可提高21.5%。
The motion estimation algorithm based on gray projection is revised to improve the image stabilization accuracy while the images among the sequence is of low contrast and their gray value is of unicity. After being preprocessed by using the Kalman filter to remove the noise, each sampling image is divided into 4 sub-regions. The offsets of each sub-region in the x and y axis are calculated, where the gray projection of the pixels in each sub-region is carried out according to the original method. When the overall offset is averaged out, the motion compensation can be payed by using the interpolation method. Based on two adjacent frames of the image sequence selected from the image database of Kyushu University, the simulation results show that the revised algorithm can improve the peak signal to noise ratio by 21.5%.
出处
《西安邮电大学学报》
2017年第2期60-64,共5页
Journal of Xi’an University of Posts and Telecommunications
基金
陕西省教育厅科学研究计划资助项目(15JK1652)
西安邮电大学研究生创新基金资助项目(CXL2015-15)
关键词
电子稳像
运动估计
灰度投影法
峰值信噪比
electronic image stabilization, motion estimate, gray projection algorithm, peak signal to noise ratio （PSNR）
