A depth image layering method and system

Abstract

The invention discloses a depth image layering method and system. The method includes: obtaining the RGB image of the current frame and the depth image of the current frame; calculating the initial optical flow according to the RGB image of the current frame and the RGB image of the previous frame; the initial optical flow includes a plurality of optical flow values; using the K-means algorithm Cluster the initial optical flow to obtain K initial clusters; one initial cluster corresponds to a Gaussian model; for any optical flow value, execute the secondary clustering process to obtain the secondary clustering result of the optical flow value; according to all The secondary clustering results of the optical flow values ​​are layered on the depth map of the current frame to obtain the layered results. The invention clusters the optical flow values ​​by using the RGB image and the Gaussian model, thereby realizing the layering of the depth image and improving the accuracy of the layering of the depth image.

Description

technical field [0001] The present invention relates to the technical field of image sequence scene flow, in particular to a depth image layering method and system. Background technique [0002] Scene Flow is essentially a three-dimensional motion vector formed when a scene or object moves. By studying the scene flow computing technology, the spatial structure information of the scene or the three-dimensional motion shape of the object can be recovered from the image sequence, so the 3D scene flow computing technology has gradually become a research hotspot in the fields of image processing, computer vision and artificial intelligence. The results are widely used in military industry, restoration of cultural relics, aerospace, medical image tracking and recognition, virtual reality and other fields. [0003] At present, the depth layering method is the most commonly used scene flow computing technology in the RGBD image sequence scene flow computing technology, while the tr...

AI Technical Summary

Problems solved by technology

[0003] At present, the depth layering method is the most commonly used scene flow computing technology in the RGBD image sequence scene flow computing technology, while the traditional depth layering method only relies on the depth image captured by the depth map camera. The distance of the class center is used to layer the depth map. The depth information captured by the depth map camera is inaccurate, and the obtained depth layering results have large errors. Especially for complex scenes such as occlusion and large displacement in the image sequence, it is easy to produce Problems such as excessive layering and edge blurring lead to poor accuracy of final scene flow calculation results

Images