基于像素值梯度变化的深度图修复算法

Depth Map Restoration Algorithm Based on Gradient Change of Pixel Value

深度图修复是计算机视觉研究领域一项重要研究问题。当前的深度传感器,在采集深度的时候会出现空洞以及物体边缘模糊化问题。为此,提出一种基于像素值梯度变化的深度图修复算法。该算法首先结合灰度图将深度图中含有空洞的区域分为两部分,然后对不同部分的像素点对应的深度值采用不同的方法进行修复:①针对空洞为灰度图像素梯度变化较低的部分直接利用邻域内的有效深度值,计算空洞区域内像素的深度值;②针对空洞为灰度图像素梯度变化较大的部分将像素点邻域内与该像素点权值系数最大的点所对应的深度值作为该点的深度值。

Depth map restoration is an important research task in the field of computer vision.The problems of void and edge blurring often appear when a depth sensor collects depth information.To address this problem,proposes a depth map restoration algorithm based on the gradient change of pixel value.First,divides the areas with voids in the depth map into two parts based on grayscale image,then proposes two differ?ent strategies to repair these two areas,respectively:1)For the part with low pixel gradient variation,directly uses the effective depth value in the neighborhood to calculate the depth value of pixels in the void region;2)For the parts with large changes in pixel gradient,the sur?rounding pixel points are weighted by distance,and the depth value corresponding to the point with the maximum weight coefficient of the pixel point is taken as the depth value of the point.