Depth image restoration method based on improved FMM algorithm
Lin Li, Huaiyu Wu, Zhihuan Chen · 2021
Aiming at the lack of depth information obtained by Kinect sensor, the collected depth images have holes, noise that affect subsequent practical applications such as the accuracy of semantic map construction , proposed a method of improved FMM algorithm to improve the quality of depth images. Firstly, clustering color images by K-means, the result as a guide image; Secondly, according to the source of the holes, it is divided into object surface holes and occlusion holes; The holes introduced by the surface holes of the object are repaired by the improved Fast Marching Method algorithm(FMM), the occlusion holes are filled by the Directional Joint Bilateral Filter(DJBF), and finally the adaptive median filter is used to denoise. The experimental results show that the way can effectively repair holes and the edge details of the depth map are clearer, and more importantly, the restored point cloud image is more complete.