一种基于基元匹配的立体图像修复算法

针对立体图像修复中直接采用二维图像修复算法所导致的深度信息修复效果欠佳的问题,提出一种基于基元匹配的立体图像修复算法.通过在立体图像左右视图间进行基元匹配,求得各个像素对应目标的深度信息;然后通过平滑深度求得缺损区域的深度数据;最后通过立体图像成像公式计算出左视图中用于填补右视图缺损区域的对应像素点坐标.通过与Criminisi算法进行对比实验,结果表明,对于多幅缺损的立体图像,该算法均取得了较好的修复效果.

A 3D Image Completion Algorithm based on Matching Primitive Elements

Aiming at the deficient of depth information repair resulted by applying 2D image completion algorithm to 3D image completion directly, a 3D image completion algorithm based on matching primitive dements was put forward. Firstly, the responding depth information of each pixel can be obtained by matching primitive elements between left and right view. Then, the depth information of defect area can be got by smoothing the depth. The responding pixel coordination in left view obtained by stereo imaging formula can be finally used to fill up the defect area in right view. Through the contrast experiment with Criminisi algorithm in several defect stereo images, the proposed algorithm shows better repairing effects.