基于图像边缘特征检测的单目立体视觉算法

Monocular Stereo Vision Algorithm Based on Image Edge Feature Detection

针对立体图像视觉技术存在仿真制作周期时间长、成本高、缺少有效的单摄像头图像采集进行立体视觉重建算法等问题,基于图像分割、傅里叶变换、霍夫变换、仿射变换,提出一种新的基于图像边缘特征检测的单目立体视觉算法。该算法首先将输入的图像进行矫正,然后对矫正后的图像进行分割,得到图像边缘特征信息;然后将分割的图像进行处理,再通过交互系统进行相关坐标点的选定,将点的仿射变换运用到整个图像;接着根据图像的边缘特征信息得到一个原图像近似的阴影,再对阴影进行相关处理;最后和原图像融合,最终使图像中形成物体类似的阴影,达到使图像立体化的效果。实验结果表明,该算法能够对不同复杂度的图像进行处理,不仅能生成单个物体的阴影,也可以同时生成多个物体的阴影,使图像能产生立体化的效果。

In order to solve the problem of stereo vision technology,such as long simulation production cycle,high cost and lack of effective single camera image acquisition algorithm for stereo vision reconstruction,a new monocular stereo vision algorithm of image edge feature detection based on image segmentation,Fourier transform,Hough transform,affine transformation and other algorithms is proposed.The algorithm firstly corrects the input image and then segments the corrected image to obtain the edge feature information.Then the segmented image is processed,and the relevant coordinate points are selected through the interactive system,and the affine transformation of points is applied to the whole image.An approximate shadow of the original image is obtained according to the edge feature information of the image,and then the shadow is correlatively processed.Finally,the image is fused with the original image,and the shadow similar to the object is formed in the image to achieve the effect of three-dimensional image.The experiment shows that the proposed method can process images with different complexity,not only can generate shadows of a single object,but also can generate shadows of multiple objects at the same time,so that the image can produce three-dimensional effect.