A multi-feature-based superpixel clustering method and device

Abstract

The present invention provides a multi-feature-based superpixel clustering method and device. The method fuses image boundary features and texture features, and realizes superpixel segmentation in combination with various features of the image. Through the clustering phase and the merging phase. In the clustering stage, the belonging of pixels is determined by measuring the similarity between pixels, and the pre-segmented superpixels are obtained by clustering; the similarity measurement in the clustering stage starts with the color characteristics and spatial position characteristics of pixels, and adds boundary feature factors to reconcile The degree of similarity between pixels located in the vicinity of the boundary. The edges of the superpixels obtained by the process closely fit, but the segmentation is too fine, so it needs to be further corrected. In the merging stage, the degree of similarity between superpixels will be measured according to the uniqueness of the superpixel content, and the scattered superpixels will be aggregated to obtain the final superpixel. Compared with the existing method, the present invention shows higher performance in maintaining the fit between the superpixel boundary and the object boundary in the image.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a multi-feature-based superpixel clustering method and equipment. Background technique [0002] Clustering algorithm, also known as group analysis, is a statistical analysis method for studying sample or index classification problems, and it is also an important algorithm for data mining. A clustering algorithm is composed of several patterns, usually, a pattern is a vector of measures, or it may be a point in space. [0003] In the technical field of image processing, clustering algorithms often only use color distance and spatial distance when measuring the similarity of pixels, so the generated superpixels generally have low edge fit, and the segmentation quality of image boundary regions is low. , cannot meet the needs of use. Contents of the invention [0004] In order to overcome the deficiencies in the above-mentioned prior art, the present invention provides ...

AI Technical Summary

Problems solved by technology

[0003] In the technical field of image processing, clustering algorithms often only use color distance and spatial distance when measuring the similarity of pixels, so the generated superpixels generally have low edge fit, and the segmentation quality of image boundary regions is low. , cannot meet the needs of use

Images