视频消旋系统中改进的Harris角点提取算法研究

Research on Improved Harris Corner Detection Algorithm for Video Eliminating Rotation System

针对传统Harris角点检测方法重合度低,不能满足视频消旋系统需求这一问题,本文提出了一种改进的Harris角点提取方法。首先将原来的梯度算子改为Prewitt差分模板,增强了Harris角点算子的旋转不变性,提高了Harris角点提取算法的重合度。同时改变传统Harris角点提取方法中的滤波窗口,减少了特征点提取的计算量。实验结果表明:该方法将传统Harris角点提取方法的吻合度提高了11%左右,提取到的图像角点不仅精确度高,且伪角点少。运算时间降为原来的30%左右,能够更好地选取特征区域,从而满足视频电子消旋算法对旋转不变性和实时性的需求。

According to the traditional Harris corner detection method has low coincide degree and can’t meet the re-quirement of video eliminating rotation, an improved Harris corner detection method is proposed in this paper. The method enhances the rotation invariant and improves the coincidence degree of corner extraction by changing the gradi-ent operator to the Prewitt mask. Moreover, the smooth filtering window of the traditional method is changed which greatly reducing the operand of detecting feature points. A great deal experiment results indicate that the consistence de-gree using the new Harris corner detection method is increased about 11% comparing to the traditional method,and the algorithm can detect corner with high accuracy and less pseudo-corner which can detect the feature area better. The processing time is decreased to about 30% of the original method,which meet the demand of rotation invariant and re-al-time for video eliminate rotation system better.