基于最小卡方统计的刑侦图像目标提取算法

An object extraction algorithm for forensic images based on minimum chi-square statistics

针对刑侦图像的自动目标提取问题,提出一种基于最小卡方统计的图像阈值化分割算法。该算法在图像一维直方图的基础上,利用卡方统计计算原图像与二值化后图像之间的相似性程度,从而获取最佳的分割阈值,提取图像中的目标区域。在实际刑侦图像库上的测试结果表明,该算法可以有效提取不同类型刑侦图像中的目标区域。

To extract objects in forensic images,a minimumχ2 statistics image thresholding segmentation algorithm is proposed.In this algorithm,the best threshold is obtained by extracting the object regions from the image and by calculating the minimumχ2 statistics between the original image and the threshold image based on the one-dimensional histogram of the image.Experiment results on the actual forensic image database show that the proposed method can extract the object regions from the different forensic images effectively.