红外序列图像清晰度评价对热像仪测温标定至关重要。文中针对传统空域评价算法的不足，提出一种基于图像轮廓线分割的红外序列图像清晰度评价方法。此算法基于变分水平集理论以能量泛函的最小化为目标，来求解使曲线趋向于目标，从而提取出红外图像中目标物的轮廓线；通过选取目标物体轮廓线两侧灰度均值大小来描述图像的清晰程度。以红外热像仪在标定时获取的序列图像为研究对象，运用文中所提出的方法与SMD，EOG，Robert，Wavelet 算法之间的对比性分析。实验表明：文中所提方法满足清晰度评价函数所具有的单调性，无偏性与良好的灵敏度。

Evaluation of an infrared image’s sharpness plays an important role in the process of calibrating the temperature of an infrared thermal imager. Aimed at increasing the efficiency of the existing estimated function of clarity, this paper puts forward an algorithm to evaluate the spatial resolution and is based on the segmentation of contour line of the image. The algorithm is based on the variational level set theory, which can move the curve toward target by solving the minimization of energy functional; subsequently, we can extract the contour line of the target from the picture and also judge the clarity of the image by subtracting the average grayscale values from the two sides of the contour line. We treat the image, which is obtained in the process of calibration by the infrared thermal imager, as target and perform comparisons among: the method which we propose in this article, SMD, EOG, Robert, and Wavelet. Experimental results show that the method depicted in this article can better meet the monotonicity and unbiased requirements, and it has higher sensitivity to the sharpness evaluation function.