Unlock instant, AI-driven research and patent intelligence for your innovation.

Picture matting method based on codec network and guide graph

A codec and guide map technology, applied in the field of image processing, can solve the problem of not being able to use sketches or click pictures, etc., to reduce the difficulty of input and improve the effect of robustness

Pending Publication Date: 2021-12-24
SHANGHAI UNIV
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] Aiming at the deficiency that the existing ternary map-based cutout technology cannot use sketches or click pictures to realize cutouts, the present invention proposes a cutout method based on a codec network and a guide map, through ternary pictures, sketches, clicks Image or all-gray input as a guide image can be accurately cut out through simple operations

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Picture matting method based on codec network and guide graph
  • Picture matting method based on codec network and guide graph
  • Picture matting method based on codec network and guide graph

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0019] Such as Figure 4a As shown, it is a method of matting based on the codec network and the guide map involved in this embodiment. Draw a guide map according to the original image and perform the first prediction through the codec network to obtain the foreground mask. According to the prediction The foreground mask of the original guide image is modified and predicted again by the codec network, and the cycle repeats until an accurate matting result is obtained.

[0020] Such as figure 1 As shown, the foreground mask is obtained by using the codec network of this embodiment according to the feature map formed by connecting the RGB image and the guide map in the channel dimension.

[0021] Such as image 3 , Figure 5 , Figure 6 As shown, the codec network includes: an encoder part, a semantic information fusion module, three skip connection modules and a decoder part including a spatial attention module, wherein: three skip connection modules are respectively arrang...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention provides a picture matting method based on a codec network and a guide graph. The method includes the steps of drawing a guide graph according to the original graph, and performing first prediction through a codec network to obtain a foreground mask; and modifying the original guide map according to the predicted foreground mask, predicting again by the codec network, and repeating until an accurate matting result is obtained. According to the method, accurate matting can be realized through simple operation by taking a ternary graph, a sketch, a click graph or full-grey input as a guide graph.

Description

technical field [0001] The present invention relates to a technology in the field of image processing, in particular to a method for matting a picture based on a codec network and a guide map. Background technique [0002] Image matting is to generate a foreground mask through the input image to separate the foreground object (the object to be extracted) from the background. The general matting problem is modeled as solving the formula I i = α i f i +(1-α i )B i , where α∈[0,1], I represents the input image (three channels), α (single channel) represents the foreground mask, F and B (three channels) represent the foreground and background regions respectively, and i represents the is the position of each pixel; in order to solve this equation, it is necessary to introduce additional constraints, general constraints include ternary maps (trimap) or sketches (scribbles). Current matting methods are mainly based on deep learning methods, and the accuracy of these methods i...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Applications(China)
IPC IPC(8): G06T7/194G06K9/62G06T7/62
CPCG06T7/194G06T7/62G06T2207/10004G06T2207/20016G06F18/25
Inventor 程航徐树公
Owner SHANGHAI UNIV