Character recognition method and system

Abstract

The invention discloses a character recognition method and a system thereof. The method comprises: obtaining geometric characteristic parameters required by correcting a distorted image; equidistantly plotting grids on the distorted image, projecting grid points to an imaging plane; processing quadrangle transformation on each grid to obtain the corrected images of the grids, assembling the corrected images of all grids into a final corrected image; determining the regions of an object character field, matching the region to a region on the final corrected image to do pretreatment, dividing to obtain independent characters; using a recognition algorithm based on gray to recognize independent character images. The technical scheme can recover distorted image to initial state, extract image characters and recognize distorted image.

Description

technical field

[0001] The present invention relates to the technical field of image processing, in particular to a method and system for character recognition. Background technique

[0002] At present, optical character recognition (OCR) technology has been very mature and formed a series of products, among which Chinese OCR recognition technology has been widely used in production and life, and the recognition rate of Chinese printed characters can reach more than 98%. However, the current OCR technology can only process high-definition images obtained from scanners. The images obtained from scanners are obtained under the condition that the source documents are completely flat, so there is almost no geometric distortion, and the obtained image values ​​may be in the There is a certain inclination in the angle, and the correction processing for this inclination is actually quite mature, and almost all commercial OCR software or document acquisition software for scanners ha...

Images