A method for extracting depth-continuous object images from images

Abstract

The invention relates to a method for extracting depth continuity object images from an image. The method is particularly characterized by comprising steps that, 1, a depth image standing for a distance from an object to a camera is acquired; 2, a first total-graph scanning is carried out, depth continuity relation among pixel points are determined, first mark operation for marking depth continuity relation of the pixel points on a labeled graph is carried out; 3, during a process of first total-graph scanning, mark relation table operation on mark numbers of the depth image is carried out; 4, final mark number integration operation of a mark relation table is carried out; 5, second total-graph scanning on labeled graphs is carried out to realize merging mark, after second mark operation, the pixel points of a depth continuity object stand for same mark number in a mark graph; and 6, the depth continuity object mark graph is outputted. Through the method, stable and effective segmentation of the target object image can be realized through depth continuity of the target object in the depth image, and thereby image processing efficiency is greatly improved.

Description

technical field [0001] The invention belongs to the technical field of computer image processing, in particular to a method for extracting depth continuous object images from images. Background technique [0002] At present, for the target object to be segmented in the image, the color map is mainly used for shooting and segmentation. Since the color map is affected by light and texture color, and the color map has three channels of image information, the object segmentation in the image segmentation is not accurate. Stability and low efficiency of image processing, and there has been no stable and effective solution to these problems. [0003] U.S. patent application US8625897 discloses a foreground and background image segmentation scheme (FOREGROUNDAND BACKGROUND IMAGE SEGMENTATION). Object segmentation and recognition. However, the following obvious deficiencies still exist. The real-time performance of image segmentation is poor, and image processing equipment with po...

AI Technical Summary

Problems solved by technology

[0002] At present, for the target object to be segmented in the image, the color map is mainly used for shooting and segmentation. Since the color map is affected by the light and texture color, and the color map has three channels of image information, the object segmentation in the image segmentation is not accurate. Stability and low efficiency of image processing, and there has been no stable and effective solution to these problems

Although this method can solve the problem of depth information extraction of existing objects, it is not practical in actual scenes, affects the final three-dimensional effect and cannot directly reflect the depth changes in local parts when the object moves locally; and it also has the following obvious deficiencies, This patent uses key points to calculate other depth value information, but the key points are not easy to obtain, and it is not easy to implement parallel algorithms

Images