Text Location In Scene Images

With the rapid development of computer, multimedia and communicationtechnology, a significant amount of text information increasingly appear in the formof image. The text information in the image is the key to describing andunderstanding the scenario content.This paper use the related technologies and methods of home and abroad forreference.Firstly,the paper introduce text area characteristics of jumping intensivlyand change rapidly in the horizontal line. Fourier-Laplace is used to eliminate thelow-frequency data on the basis of the characteristics.Then, on this basis, the analysis on the skeleton, they are classified into simpleand complex component, the simple components were considered to be the simpletext area, and then complex components are processed.Secondly, if there is cross point in the thin lines, the complex thin line was splitinto simple line on the cross point. So each simple thin line was thought to be a textarea. But some of the whole text region is divided into a plurality of regions.We needto merge them according to the text direction.But some simple non-text line issimilar to text line.According to the text of the edge density,the total non-text linelength is smaller than total text line length. Finally, these possible non-text area wereeliminated by the feature which a text area edge density is larger.Finally, a brief introduction on the development tools and the experimentalresults are analyzed. Experiments show that this method improves textlocation accuracy to a certain extent.