基于边缘信息引导滤波的深度图像增强算法

DEPTH IMAGE ENHANCEMENT ALGORITHM BASED ON EDGE INFORMATION GUIDANCE FILTER

由于Kinect传感器的有效测距有限,深度图像会出现黑洞、噪声等问题,传统的深度图像增强算法仅利用彩色图像信息填充黑洞,增强后的深度图像物体边缘不清晰。针对这种情况,提出基于边缘信息引导滤波的深度图像增强算法。首先,分别获取彩色图像和深度图像的边缘信息,通过融合得到作为引导的边缘图像信息;然后,将边缘信息和迭代非局部中值滤波算法相结合进行黑洞填充;最后,利用自适应中值滤波对图像进行平滑噪声的处理。实验结果表明,该算法能够很好地修复深度图像,得到较为清晰的物体边缘。

Because of l imited distance measurement of Kinect sensor, black holes and noise etc. would occur in the depth map. Previous approaches to fill the black holes only used color images information and might cause unclear object edges of depth map after strengthening. So we proposed a new depth map enhancement algorithm based on edge information guided filtering. First, we combined the edge information of the color image and depth map as guide edge image information. Second, we combined the edge information and the non-local iterative median filtering algorithm to cover the hole. Finally, we used the adaptive median filter to smooth the noise. Experimental results show that the proposed approach can produce high quality depth map and can get a sharp edge.