改进的Harris角点文本定位方法

针对计算加权角点饱和度方法在文本定位中对非文本角点提取过密或文本角点提取不足的情况容易错误定位的问题,通过运用候选区域块的高宽比(外接矩形高度与宽度之比)、面积比(白色区域面积与外接矩形面积之比)数据作为特征,结合二元线性回归提出了文本定位方程,解决非文本角点过密的问题.运用文本区域块的高宽比比面积比的数值通常出现两极分布的特点,在定位方程中加入常数C1,用来定位文本,数值在两极之间时加入常数C2,定位非文本,实验结果表明本文算法比计算加权角点饱和度方法取得了更好的效果.

Improved Harris Corner Text Location

The calculation of weighted corner saturation gets the wrong location in text when the corners are over crowding, or the corners are insufficient. This paper solves the overcrowding of non-text corner by using the aspect ratio data (the ratio of exernal rectangular height and width), area ratio data (the ratio of white area and external rectangular area) of the candidate block as the eigenvalue.This text location equation is put forward with the combination of Binary linear regression. Using the numerical of regional block ratio area of high to width ratio of the text region is usually appear the poles distribution, it employs constant C1 in the locationequation to locate the text. Then, it adds constant C2 when value between the poles to locate the non-text. The experimental results show that this algorithm has achieved better effect than the method of calculating weighted corner saturation.