基于轴对称窗口文字图像检测快速算法研究

New Algorithm Based on Symmetrical Window Text Image Detection Algorithm

研究了图像定位的问题,由于存在污渍干扰等影响图像定位,针对以往单一特征进行图像文字定位及识别的系统中容易受到各种环境因素干扰的缺陷,提出了一种利用轴对称窗口进行边缘检测的图像文字检测算法。首先将利用轴对称滑动窗口提取水平和竖直方向上的边缘特征,利用连通域确定初始的图像文字位置;通过对可能的图像文字区域进行颜色色调验证,区域内垂直方向直方图投影,从而确定最终的图像文字位置。由于利用多种特征综合检测图像文字进行仿真。仿真结果表明改进方法能准确检测出复杂场景下图像文字所在区域。

In the past,the image text location and identification system based on single feature was susceptible to a variety of environmental factors.The paper proposed an image text detection algorithm based on symmetric window edge detection.Firstly,symmetric sliding window was used to extract the edge features of horizontal and vertical.Then the connected domain was used to determine the initial image text position.The hue verification was carried out with the possible image text regions,and the histogram projection was done in vertical direction,thereby determining the final image text position.Experimental results show that this method can accurately detect the complex scene images and text area.