摘要
机器视觉是机器人导航的基础,双目立体成像是机器视觉常用的方法。双目立体成像中因为对焦的问题,生成的影像有着一定的区别,这种区别对立体成像有着严重的制约。提出一种针对多聚焦图像的拉普拉斯金字塔融合方法,左右聚焦的图像经过多尺度分解,得到不同频率域的图像细节;在对应的尺度上进行融合,得到一个融合后的多尺度图像;再进行拉普拉斯金字塔逆变换,得到融合图像。该方法有效地解决了多聚焦图像局部不清晰的问题,有利于双目视觉立体成像。
Machine vision is the basis of robot navigation. Binocular stereoscopic imaging is a common method of machine vision.In binocular stereoscopic imaging, because of the problem of focusing, there are certain differences between the generated images.This distinction has serious constraints on stereoscopic imaging. A Laplacian Pyramid fusion method for multi focus image is proposed. After multi-scale decomposition of left-right focusing images, the details of images in different frequency domains can be obtained. Then, fusion can be carried out on corresponding scales to obtain a multi-scale image after fusion. Then, inverse transform of Laplacian Pyramid can be carried out to obtain the fused image. Thus, the problem of local indistinct image in multi focus image is solved by the method, which is beneficial to binocular stereo vision imaging.
引文
[1]涂超平,肖进胜,杜康华,易本顺.基于各向异性热扩散方程的多聚焦图像融合算法[J].电子学报,2015.43(6):1192-1199
[2]潘永胜,冀小平.一种新的多特征融合图像检索方法[J].科学技术与工程,2014.14(15):219-223
[3]邓磊,李京,陈云浩,等.几种小波融合方法在遥感影像融合中的应用与比较.遥感应用,2007.36(6):23-27
[4]牛颖超,周忠发,谢雅婷,崔亮.基于金字塔变换算法优化的遥感图像融合[J].激光与光电子学进展,2017.54(1):276-283
[5]伊力哈木·亚尔买买提,谢丽蓉,孔军.基于PCA变换与小波变换的遥感图像融合方法[J].红外与激光工程,2014.43(7):2335-2340
[6]朱明,孙继刚,梁伟,郭立强.四元数曲波变换多源多聚焦彩色图像融合[J].光学精密工程,2013.21(10):2671-2678
[7]刘佳,段先华,孙庆国.基于多小波的高分辨率与多光谱图像融合[J].电子设计工程,2015.23(11):82-86
[8]阎昆,李心怡,王珺.结构稀疏表示与细节注入的遥感图像融合方法[J].电子设计工程,2018,26(06):68-71,76
[9]罗鸿斌.基于拉普拉斯高斯金字塔的图像融合研究[J].自动化与仪器仪表,2014.3:27-30
[10]张闯,常建华,葛益娴,孙冬娇.多聚焦图像的自适应分块融合方法[J].科学技术与工程,2013.13(21):6124-6128
[11]史敏红,高媛,秦品乐,王丽芳.基于NSST域的改进加权非负矩阵分解的图像融合[J].科学技术与工程,2018.18(3):268-273
[12]夏文栋,陈德礼,任江涛,刘远峰.多分辨率分析耦合近似稀疏表示的图像融合算法[J].科学技术与工程,2017.17(33):297-303