基于模型的FLIR图像的多分辨率分割方法

针对低对比度前视红外（ＦＬＩＲ）图像分割问题的特点，将多分辨率图像处理技术与基于模型的前视红外图像分割方法结合起来，提出了基于模型的ＦＬＩＲ图像的多分辨率分割方法．首先建立一金字塔结构的多分辨率图像集，然后在每一固定分辨率级别上使用基于模型的ＦＬＩＲ图像分割方法，并用低分辨率上的分割结果修正高分辨率上迭代的初始概率，最后在原ＦＬＩＲ图像中得到分割结果．该方法能得到更为精确的目标分割结果，并大大减小背景对分割图像的干扰，同时还加快了选代的收敛速度．对实际红外目标图像的分割结果证实了本算法的有效性．

Model-Based Multiresolution Segmentation of FLIR Images

n accordance with the feature of low-contrast FLIR images, the authors present a segmentation algorithm for forward-looking infrared (FLIR) images, which incorporates the model-based segmentation of FLIR images with the multiresolution image processing technique. A multiresolution image set of a pyramid like structure is derived, and then the model-based segmentation at every fixed resolution level is applied. The segmentation result at a lower resolution is used to modify the initial probability assignment at a higher resolution. The segmentation result at the original FLIR image is obtained. This algorithm will be able to give a more accurate segmentation of a target while greatly reducing the influence of the background on image segmentation, and speeding up the convergence rate of iteration. The effectiveness of the algorithm has been proved by segmentation results with real infrared images.