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

Hand writing erasing method based on deep learning

A technology of deep learning and handwriting, applied in the field of handwriting erasure, can solve the problems of restoring printed words, unsatisfactory effect, high labor cost, etc., and achieve the effects of weakening the influence, enhancing the generation effect, and strengthening the feature extraction

Pending Publication Date: 2022-07-05
NANJING UNIV +1
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

When distinguishing between handwritten and printed content, the texture and grayscale of the two are similar, and the traditional methods of locating handwritten characters such as edge detection and color positioning are invalid; when filling the handwritten area, the existing method of randomly sampling background pixel values ​​​​for filling It is impossible to fully restore the printed words when the handwritten words overlap with the printed words
Obviously, the effect of using existing methods to automatically erase handwritten characters in document images is not satisfactory, and if image editing software is used to process handwritten characters pixel by pixel, the labor cost required is too high

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
  • Hand writing erasing method based on deep learning
  • Hand writing erasing method based on deep learning
  • Hand writing erasing method based on deep learning

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0083] A deep learning-based handwriting erasing method, such as figure 1 As shown, this example is a fully convolutional neural network model, which consists of three modules: a mask generation module, a first-stage image generation module, and a second-stage image generation module. Among them, each module adopts an encoder-decoder structure, and the encoder of the mask generation module shares parameters with the encoder of the first-stage image generation module.

[0084] This embodiment provides a deep learning-based handwriting erasure model training method, and the training process is as follows figure 2 shown, the process is as follows:

[0085] 1. Create training samples

[0086] Prepare a paper handwritten document with printed content. After adding handwriting appropriately, use the scanner to obtain the document image as the original image; use Photoshop software to identify and fill the handwritten area by pixel to obtain the target image; use the algorithm to ...

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 discloses a handwriting erasing method based on deep learning. According to the method, handwriting erasing is carried out on a document image, a convolutional neural network model is used, details of printed characters and background images are reserved, meanwhile, a handwriting region is recognized, and appropriate pixel values are automatically filled. Wherein the convolutional neural network model introduces jump connection, network shallow layer features and deep semantic information are combined, and the image detail generation effect is enhanced; a deformable convolution method is adopted, so that the network adaptively adjusts the convolution sampling position, and the erasing effect of handwritten writings with different shapes and sizes is improved; through an attention mechanism, a network is guided to pay attention to feature extraction of a handwriting region, the capability of distinguishing handwriting and printing content is improved, and the influence of a complex background on an erasing effect is weakened.

Description

technical field [0001] The invention relates to a method for erasing handwritten characters, in particular to a method for erasing handwritten characters based on deep learning. Background technique [0002] Taking pictures of documents and erasing handwritten words in document images is a document restoration technology, which has a wide range of applications in office, learning and other fields, such as: (1) to protect handwritten private information in documents; (2) as preprocessing To improve the accuracy of OCR technology to recognize printed words; (3) to restore the original information of the document, such as the collection of students' wrong questions for repeated practice, the extraction of original forms for information refilling, and the removal of notes or graffiti on book images. Comparing the images before and after erasing handwritten characters can also be used for handwritten character extraction, handwritten information recognition, etc. [0003] The re...

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
IPC IPC(8): G06V30/412G06K9/62G06N3/04G06N3/08G06V10/774G06V10/82
CPCG06N3/08G06N3/045G06F18/214
Inventor 陈力军刘佳赖慧慧陈星宇鄢伟
Owner NANJING UNIV