针对气动光学效应以及噪声等先验信息不明带来的图像模糊退化, 从而使图像过渡区域不易进行细化的问题, 提出了一种大气湍流退化图像的过渡区域提取及细化算法。首先, 对退化图像进行了基于局部方差的区域分割；然后, 对过渡区域进行噪声疑似度和边缘疑似度判断, 以获得更贴近区域中心线的较窄区域；最后, 进一步对过渡区域跟踪细化完成了轮廓提取。大量仿真实验结果表明, 算法较好地提取了退化图像的轮廓信息, 同时, 算法具有较好的区域分割效果、抗噪性能、细化效果以及实时性。

It is difficult to refine edge transitional region for aero-optical degraded images due to aero-optical effects and unknown of transcendental message. To solve the problem we proposed an extracting and refining algorithm of edge transitional region for turbulence degraded images. Firstly adaptive region segmentation based on local variance was made to get edge transitional region. Secondly similar degree of noise and similar degree of edge were calculated for obtaining the narrow region close to heartland of edge transitional region. At last track refining to edge transitional region was implemented and the edge contour was extracted. The simulation result indicates that this algorithm which can extract well the contour message of turbulence degraded images has fine performance on region segmentation denoising refining and real-time capability.