An Optimal Recognition Method for Image Segmentation Based on Edge Completeness

Abstract

An edge completeness-based optimal identification method for image segmentation, comprising the following steps: performing smooth filtering on an image; obtaining an image edge point using edge detection; obtaining an initial over-segmentation image spot set by applying a segmentation algorithm; performing seed image spot identification and marking for the image spots; calculating edge completeness of the image spots, and sorting according to internal edge points and spectral difference; selecting preferential image spots to perform regional growth, calculating and storing any combination results and edge completeness, and forming an edge completeness curve; calculating a maximum value of the edge completeness curve of the image spots, and obtaining an image spot corresponding to a maximum value point; and marking the image spot corresponding to the maximum value as an optimal segmentation image object, and marking an initial over-segmentation image spot contained in an icon as a processed image spot. A segmentation result obtained by applying said method and image spot space information obtained by calculation according to the result reflect real space information of a ground object in an image, and a foundation is provided for application of space information, topological information and context information of the image in subsequent remote sensing information extraction.

Description

technical field [0001] The present invention relates to image optimal segmentation and recognition method of image object in image analysis, especially relates to a kind of integrated utilization of multiple image clues - image area and edge, continuity and discontinuity, to distinguish and recognize the segmentation result method. Background technique [0002] Image segmentation classifies similar connected pixels in the image into the same image area, which is an expression of image continuity, while image edges only reflect local differences in the image. How to judge the quality of segmentation results is an important content in image segmentation, and the quality of image segmentation has a key impact on subsequent image processing, semantic cognition and image understanding. [0003] Image segmentation provides a new solution to solve the bottleneck encountered in the data processing of spatial high-resolution remote sensing images (hereinafter referred to as "high-re...

AI Technical Summary

Problems solved by technology

Edge-based and region-based segmentation actually start from different angles of the same point, and each has its own advantages and limitations: the edge-based segmentation method detects local discontinuous pixels through the edge, and has a good detection of local boundary information. effect, but the global segmentation ability is not enough, and it is difficult to obtain closed edges; while the region-based segmentation method uses the gray level statistical information of the pixel to create a region, which can overcome the influence of noise, but at the same time, the accuracy of the specific location of the edge is not enough

Images