基于Hough变换的红头文件识别算法

Red Heading Documents Recognition Algorithm Based on Hough Transform

直线检测一直是图像处理中重要的一个步骤。最常用的直线检测技术是Hough变换,它将图像空间检测直线的问题转换到参数空间的峰值检测问题,图像空间中每一个非零像素都会对参数空间中与其对应的累加器投上一票。投票过程中,远离峰值的大多数的投票都是不必要的,这些无效投票降低了转换速度。提出一种在图像空间中基于指定的短直线形态过滤掉指定方向之外的多余线条和大部分噪声的方法,降低投票量,并且还可以将投票的范围限定在指定范围内。实验结果证实了该方法的有效性。

Line detection has always been an important step in image processing. The most commonly used line detection technique is the Hough transform,which converts the problem of image space detection line to the peak detection problem in the parameter space. Each non-zero pixel in the image space will vote for its corresponding accumulator in the parameter space. During the voting process,most of the votes that are far from the peak are unnecessary,and these invalid votes reduce the conversion speed. A method for filtering out excess lines and most noise outside the specified direction based on the specified short line form in the image space is proposed,the voting amount is reduced,and the voting range can be limited to the specified range. The experimental results confirm the effectiveness of the method.