基于异形曲面拟合引导的深度信息修复算法

DEPTH INFORMATION RESTORATION ALGORITHM BASED ON FITTING GUIDANCE OF SPECIAL-SHAPED SURFACE

为了克服异形投影曲面RGB-D空间的深度信息出现非预期抖动、表面采样缺失等问题,提出一种基于异形曲面拟合引导的深度信息修复算法。通过将格雷码序列图像投射于异形投影曲面对其赋予高可用纹理特征,使纹理形态为原始深度信息的修复过程提供强约束引导参照,以实现异形投影曲面的非重建式平缓区域划分。再利用原始深度信息划分结果对各平缓区域分别进行曲面拟合,然后根据拟合曲面方程对该区域内原始深度信息进行平滑修复,从而增强异形投影曲面深度信息的可用性及可信性。算法验证结果表明,该深度信息修复算法不仅能够实现异形曲面的空洞填补与平滑修复,还能够有效地减少深度信息过修复、欠修复等情况发生。

In order to overcome the problems such as unexpected dither of depth information in RGB-D space and the lack of special-shaped surface sampling,a depth information restoration algorithm based on the fitting guidance of special-shaped surface is proposed.The gray code sequence images were projected on the irregular projection surface to give it highly available texture features,so that the texture shape provided strong constraint guidance reference for the repairing process of the original depth information,so as to realize the non-reconstruction smooth region division of the irregular projection surface.Then we used the original depth information partition results to fit the surface of each smooth region,and then the original depth information in the region was smoothly repaired according to the fitting surface equation,thereby enhancing the availability and credibility of the depth information of the irregular projection surface.The validation results of the algorithm show that the depth information repair algorithm can not only fill the hole and smooth the irregular surface,but also effectively reduce the over repair and under repair of the depth information.