基于深度融合的深度图像修复算法

提出了一种基于深度融合的深度图像修 复算法。对于单幅深度图像,首先利用形态学操作进行空洞区域优化,消除深度图像中的间 隙和随机噪声；然后针对迭 代滤波过程,提出一种新的深度融合策略计算深度值,并通过对空洞区域的分析,判断深度 图像中空洞区 域类型,自适应选择结构元进行迭代操作；最后利用局部深度值重建方法对受损的边缘处深 度值进行修复。 实验结果表明,本文算法在较好的修复深度图像中存在空洞和间隙的同时,能够保持原始深 度图深度值分 布规律,克服修复过程中存在的深度值失真,边缘模糊等不足。基于标准数据集Middlebury 的对比试验结果表明,本文算法与其它算法相比,获得了良好的效果。

A depth image inpainting algorithm via depth fusion

A depth map is a special image that describes the depth of field infor mation of a scene. In order to solve the problem that the depth map often has holes and gaps at the boundary of the object due to random noise and equipment performance,a depth image restoration a lgorithm is proposed based on depth fusion.For the single depth map,the morpholo gical operation is first used to optimize the cavity region to eliminate the gap and random noise in the depth map.Then,for the iterative filtering process,a new depth fusion strategy is proposed to calc ulate the depth value and pass the cavity region.The analysis determines the type of the hollow hole region of the depth image and adaptively selects the structural element to perform the ite rative operation. Finally,the local depth value reconstruction method is used to repair the depth value at the damaged edge.The experimental results show that the proposed algorithm can main tain the depth distribution of the original depth map while maintaining the depth and gap in th e better depth image,and overcome the shortcomings such as depth value distortion and edge blu r in the repair process.The comparison test results based on the standard data set Middlebury s how that the proposed algorithm achieves good results compared with other algorithms.