Method for rapidly calculating characteristic differences of texts and images

Abstract

The invention discloses a method for rapidly calculating characteristic differences of texts and images. The method includes the steps that firstly based on diversity of geometrical shape characteristics, the texts with large differences are preliminarily screened according to four characteristics of the number of foreground pixels, the sizes of the texts, the length-width ratio and the duty ratio, then, the texts are effectively described on the whole, and the negative effects caused by image distortion are reduced at the same time; after preliminary screening is conducted, fine characteristics of remaining pairs are described, a corrected template matching method is adopted, according to the corrected template matching method, a given template slides in the images, and then targets matched with the template are acquired. Coarse characteristic screening and fine template matching are sequentially conducted on text blocks, the differences of the sizes of the text blocks are limited by the sizes of the texts in coarse characteristics, and therefore the number of comparison times in the matching process is greatly reduced. Furthermore, the invention provides a lookup table associative bit memory accelerating method, and the performance of the accelerating method is improved by building the corresponding relation between the text blocks and data in advance according to similarity measurement characteristics.

Description

technical field [0001] The invention belongs to the field of computer vision and image processing, and relates to a fast calculation method for feature differences of characters and images. Background technique [0002] In the prior art, the similarity measurement method for large data volume features mainly has the following aspects: [0003] 1) Video text recognition based on wavelet transform and similarity measure, Computer Engineering and Design, 2008, Yang Qiang, Cheng Yukun, Ma Sen. [0004] The processing object of this method is subtitles in videos. It mainly studies the extraction and recognition of subtitles in videos, and uses the method of similarity measurement to identify video subtitles. First, the character image to be matched and the character image in the template library are respectively decomposed into N levels, and the low-frequency parts after each level of decomposition are reserved, and then the error accumulation method is used to roughly search fo...

AI Technical Summary

Problems solved by technology

[0011] Although the existing methods have proposed many similarity measures, there is no similarity measure for a large number of scanned text images.

Images