Abstract: No reference image quality assessment is still one of the most challenging areas in
computer vision. As forming mechanism constraints, infrared reconnaissance image quality is less
than visible light images. Low quality will severely hinder efficiency of subsequent image processing
and surveillance, an accurate assessment of image quality is the key link. To this end, we study and
extract the ability to characterize the infrared radiation characteristic contours, textures, and point
features, construct infrared image quality evaluation model based on mixing characteristics and
significance detecting. And we test the result by taking infrared reconnaissance image as a sample and
subjective evaluation as a measure of reference. For the relevant experiments, the results show that
the method is more competitive than classical method and it can be used to evaluate the quality of
infrared reconnaissance images.

Key words: infrared reconnaissance image, image quality assessment, saliency detection, mixed
features