基于窗口Hough变换与阈值分割的矩形识别算法

提出一种基于窗口霍夫变换与阈值分割自动识别图像中的矩形策略： 通过图像窗口霍夫变换，提取霍夫图像的峰值（对应原始图像的线段），当四个峰值满足某些几何条件时，则检测出矩形；对图像进行阈值分割，将分割结果与霍夫变换的矩形做拟合修正.对不同成像背景和光照环境下图像的集成测试结果表明，本策略能够很好地抑制在多种自然光照不均和拍摄角度造成的干扰.且采用了缩略图计算，降低了逐像素运算的时间复杂度，可满足实时性要求.该技术可运用在实时准确裁剪银行票据目标等各个需要快速识别矩形的工程领域.

Rectangle Detection Algorithm Based on Windowed Hough Transform and Threshold Segmentation

This study proposes a method for automatic recognition and cutting of bank bills using windowed Hough transform: by scanning each pixel to compute the Hough transform of the image and extract the peaks of the Hough transform (which correspond to line segments). A rectangle is detected when four extracted peaks satisfy certain geometric conditions (which correspond the border of bills). Threshold is used to segment the source image and perform fitting correction for the segment result and the extracted Hough transform rectangles. The integration test results of on different image backgrounds and illumination environment indicate that the proposed strategy has a good ability to suppress the interference caused by different natural illumination and shooting angles. In addition, thumbnails view is used to extract feature, reducing the time complexity of pixel-by-pixel operation.