Image level extraction and editing method based on global energy function optimization

Abstract

The invention provides an image level extraction and editing method based on global energy function optimization. The method includes the steps that an image is subjected to hyperfractionation to obtain a plurality of hyperfractionation areas, local characteristics of relevant hierarchical relations in various areas are extracted, and the foreground-background relation and the same-layer merging relation are judged through local characteristics; on the basis of the judgment result, a global energy function is established, and the function is solved with an improved differential evolution algorithm so that final hierarchical relations can be obtained; according to the hierarchical relations of the image, the partial ordering relations between all the levels are edited, and when original partial ordering relations change, covered levels in the original image are filled with contours and colors. With the method, local information and global information can be combined, and the image can be edited on the basis that the image hierarchical relations are calculated and obtained.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image layer extraction and editing method based on global energy function optimization. Background technique [0002] Image editing is the core technology in all image processing software. Editing the layers in an image is a very challenging content in image editing, which requires an understanding of the semantic information of the image. The key point is that we need to get the hierarchical relationship of an image, and then use the existing outline filling and color filling methods to edit the order and color of the image hierarchy. Extracting the hierarchical relationship of objects in an image is a key step in image scene understanding and further high-level vision tasks, including foreground / background separation, depth prediction from a single image, 2.5-dimensional parsimonious maps, image and video coding, motion analysis, etc. For example, for image coding...

AI Technical Summary

Problems solved by technology

[0003] How to extract the hierarchical relationship of the image is still a very challenging problem. The mainstream image editing software such as Photoshop does not support the automatic extraction of the hierarchical relationship. The layer information in the image needs to be manually selected by people, which is very inconvenient.

None of the existing methods for extracting image hierarchical relationships can be used as a general-purpose technology. Some methods have high complexity and are only suitable for processing simple artificial images and are not conducive to processing complex natural images; some methods only use local features. Information, can locally judge the occlusion relationship of adjacent areas but lacks the constraints of the global partial order relationship; some methods conduct research on the basis of obtaining the correct segmented area, which puts forward high requirements for the initial image segmentation results

Images