Moire text image binarization system based on graph residual attention network
An image binarization, text image technology, applied in the field of computer vision, can solve the problem that the binarization network is not suitable for processing images with moiré text, there is no moiré text image binarization system, and it is difficult to deal with large-scale moiré. problems such as texture images, to achieve the effect of convenient construction and update, avoiding gradient disappearance, and robust performance
Active Publication Date: 2019-09-27
DALIAN UNIV OF TECH
View PDF3 Cites 13 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
However, the current deep learning network may lose part of the content information while removing the moiré pattern, and because the connection between the branches of each scale is not considered, the moiré pattern is not completely removed; at the same time, the existing binary network is not suitable for processing with Images with moiré text, especially images with large-scale moiré, are difficult to handle
So far, there is no dedicated system for binarizing Moore text images
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0028] A Moore text image binarization system based on graph residual attention network of the present invention such as figure 1 As shown, the moiré removal module is connected with the image binarization module and the output features of the moiré removal module are subtracted from the output features of the image binarization module. The moiré removal module is composed of interconnected multi-scale volumes The product residual module and the triple attention module are composed, and the image binarization module includes a plurality of sequentially connected convolutional layers that generate adaptive local thresholds, a nonlinear activation function layer, and a parameterization layer;
[0029] The multi-scale convolution residual module includes 5 parallel resolution branches, and each resolution branch is composed of a sequentially connected downsampling layer, a residual module and an upsampling layer; the downsampling layer is a convolution kernel of A 3×3 convolution...
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
Login to View More Abstract
The invention discloses a Moire text image binarization system based on a graph residual attention network. The system is characterized by firstly, extracting the features with different resolutions by utilizing a multi-scale convolutional neural network, respectively removing the Moire patterns in the features with different resolutions, and keeping the content information in a text image to avoid gradient disappearance; guiding a target area in the network attention characteristics through a triple attention module by utilizing the channel information, the spatial information and the relationship among the branches, and further removing the moire patterns. The set image binarization module can accurately binarize the moire-removed text image by utilizing a plurality of convolution layers which are connected in sequence and generate an adaptive local threshold value and a parameterized layer for promoting binarization, the moire patterns in different frequency band ranges and channels can be fully removed, and the image binarization precision is improved.
Description
technical field [0001] The invention relates to the technical field of computer vision, in particular to a Moore text image binarization system based on a graph residual attention network with simple training and robust performance. Background technique [0002] Text image binarization is the basis and key step of most text analysis tasks (such as text content recognition, optical character recognition), which essentially assigns different binary values to the target text and background of the text image. In many cases, people need to use cameras and other equipment to record documents on electronic screens, and the appearance of moiré has brought new challenges to the binarization of such text images. At present, the rise of deep learning has brought breakthroughs to the image binarization and moiré removal problems, and has greatly improved the performance of the image binarization system and the moiré removal system. However, the current deep learning network may lose ...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/38G06N3/04
CPCG06V10/28G06N3/045
Inventor 郭艳卿姬彩娟郑欣
Owner DALIAN UNIV OF TECH