改进的交互式Otsu红外图像分割算法

为提高红外图像中目标分割的精度和抗噪性能,提出了一种改进的交互式Otsu图像分割算法。采用图像信息熵特征和类间方差特征对经典Otsu算法的阈值判别函数进行改进,获得的最优阈值能较好地将目标从背景中分割出来,且具有良好的边缘保持效果,提高了算法的分割精度。同时,针对红外图像目标单一的特点,采用交互式粗分割的思路,先在红外图像中提取包含目标的局部封闭区域,进而在提取的区域内进行改进的Otsu分割。通过对红外图像激光光斑目标提取过程的实验结果表明：改进的Otsu分割算法大大降低了背景噪声对分割算法的影响,提高了抗噪性能与分割精度,且最大程度地减少分割算法的运算量,并较好地保持了目标模糊边缘,分割效果优于传统的Otsu算法和相关的改进Otsu算法。

An improved interactive Otsu infrared image segmentation algorithm

To improve the precision and anti-noise performance of infrared image segmentation, an improved and interactive Otsu image segmentation algorithm is proposed. The characteristics of entropy and class variance of images are used to improve the threshold discriminant function of the classical Otsu algorithm that obtains the optimal threshold. The optimal threshold can segment the target from the background and has a good edge retention effect, which improves the segmentation accuracy of Otsu algorithm. Meanwhile, for the feature of single target in infrared image, the idea of interactive rough split is used before segmenting the image. The partial closure of the area containing the target is extracted firstly in the infrared image, and then in the extracted region. The experimental results for the extraction of laser spot target in infrared image show that the improved Otsu segmentation algorithm greatly reduces the impact of the background noise on the segmentation algorithm, also improves the anti-noise performance and segmentation accuracy of the traditional Otsu algorithm and minimizes the computation of the segmentation algorithm, maintaining the target fuzzy edge. The segmentation performance is better than the traditional Otsu segmentation and several related improved Otsu algorithms.