应用区域生长的无人船红外图像精确分割方法

Accurate Segmentation of Unmanned Ship Infrared Image Using Region Growing

在分割红外图像的过程中,受热量传递方式的影响,像素点的特征较为模糊,导致分割后各部分的结构相似性有待提升。因此,提出应用区域生长的无人船红外图像精确分割方法研究。首先,通过离散处理的方式调整原始图像像素点排布方式。其次,利用脉冲系数的波动阈值范围约束离散程度,根据离散后像素点的灰度值特征实现降噪。最后,引入了区域生长算法,将图像中亮度最高的像素点作为种子点,通过构建n×n的矩阵遍历整个图像,当遍历区域内像素的灰度均值和标准差在自适应阈值范围内时,则将其作为种子点的生长区域,以此实现对图像的分割。实验结果表明,所提方法对不同质量无人船红外图分割后,对应的结构相似性稳定在0.82以上,明显优于对比方法。

In the process of segmenting infrared images, the features of pixel points are blurred by the heat transfer method, which leads to the structural similarity of each part after segmentation to be improved. Therefore, the research of accurate segmentation method of unmanned ship infrared image applying region growing is proposed. Firstly, the original image pixel point arrangement method is adjusted by discrete processing. Secondly, the fluctuation threshold range of pulse coefficients is used to constrain the degree of dispersion, and noise reduction is realized according to the gray value characteristics of the discrete pixel points. Finally, a region growing algorithm is introduced,in which the pixel point with the highest brightness in the image is used as the seed point, and the whole image is traversed by constructing an n×n matrix, and when the gray value mean and standard deviation of the pixels in the traversed region are within the adaptive threshold range, it is used as the growing region of the seed point to realize the segmentation of the image. The experimental results show that the corresponding structural similarity of the proposed method is stable above 0.82 after segmentation of infrared images of unmanned ships of different quality, which is significantly better than the comparison method.