基于图像熵的自适应阈值配准算法  

Adaptive Threshold Registration Algorithm Based on Image Entropy

作　　者：袁俊鹏 安维胜[1] 苟鹏 YUAN Junpeng;AN Weisheng;GOU Peng(College of Mechanical Engineering,Southwest Jiaotong University,Chengdu 610031)

机构地区：[1]西南交通大学机械工程学院,成都610031

出　　处：《计算机与数字工程》2022年第6期1354-1358,1370,共6页Computer & Digital Engineering

摘　　要：针对目前主流鲁棒估计算法匹配精确率较低、速度无法满足实时性场景等问题,提出了一种基于网格运动统计(GMS)的改进方法。首先结合图像熵提出一种自适应阈值函数实现特征点全局分布;其次对网格评分模型优化为快速4-邻域网格模型,减少了计算时间。实验结果显示,论文算法在保持较高的匹配正确率的同时,较原算法在运行时间上平均降低了20%~28%,在旋转、尺度、光照等变化时具有很强的鲁棒性。In order to solve the problems that the current mainstream robust estimation algorithm has a low matching accuracy rate and the speed cannot meet real-time scenarios,an improved method based on grid motion statistics(GMS)is proposed.First,an adaptive threshold function is proposed in combination with image entropy to realize the global distribution of feature points.Secondly,the grid scoring model is optimized to a fast 4-neighbor grid model,which reduces the calculation time.Experimental results show that,while maintaining a high matching accuracy rate,the algorithm in this paper reduces the running time by an average of 20%to 28%compared with the original algorithm,and is very robust against changes in rotation,scale and lighting.

关 键 词：图像配准 网格运动统计 图像熵 自适应阈值 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]