The mesh model and its construction method

Abstract

The invention relates to the technical field of image processing, in particular to a slice-network model used for image representation and a method for constructing the slice-network model. The slice network model used for image representation is a forest structure composed of several compound nodes and basic nodes, each of the compound nodes is a non-leaf node, and each of the basic nodes is a leaf node; The basic node includes a certain area of ​​the image and a representative piece representing the apparent characteristics of the area; the composite node includes a certain area of ​​the image and can be further decomposed into basic nodes and / or composite nodes; located in the forest structure There is an edge between two nodes of the same layer and connected in space, and the edge has a relationship matrix representing the relative position of the two nodes in space.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a slice-network model used for image representation and a method for constructing the slice-network model. Background technique [0002] In recent years, various image processing techniques have become a hot field of computer science research. Among them, slice-based image synthesis and editing technology has become a powerful tool for users to edit multimedia content. [0003] Patch-based image editing methods in the prior art, typically PatchMatch proposed by Barnes et al. in 2009, require users to provide semantic guidance or interaction to obtain meaningful results; moreover, for editing based on multiple images, Users need to search for new objects or areas to synthesize to the specified position; but these require users to have strong art expertise and proficiency in software; at the same time, in the case of many candidate images, users need to spend a long time ...

AI Technical Summary

Problems solved by technology

[0003] Patch-based image editing methods in the prior art, typically PatchMatch proposed by Barnes et al. in 2009, require users to provide semantic guidance or interaction to obtain meaningful results; moreover, for editing based on multiple images, Users need to search for new objects or areas to synthesize to the specified position; but these require users to have strong art expertise and proficiency in software; at the same time, in the case of many candidate images, users need to spend a long time boring labor

The NRDC (Non-Rigid Dense Correspondence Method) proposed by HaCohen et al. in 2011 can be applied to the above-mentioned scenarios by automatically finding suitable image slices for synthesis; however, the disadvantage of this method is that the processing speed is slow and cannot meet real-time image editing. Especially for image library based editing needs

Image search methods can also be used to find suitable image sources for the above scenarios; however, image search methods are generally based on global features, and cannot find suitable and optimal partial image regions for local image editing

[0004] To sum up, on the basis of the existing image synthesis technology, the main difficulty of image editing based on slices in the image database is that the slices used in the editing technology only describe the local appearance features, and do not The image context or the description of the image itself cannot effectively organize and search the content in the image library, and the method of image search using global features lacks accurate positioning of local information

Images