Thermal infrared spectrometer on-orbit defocus assessment based on blind image blur kernel estimation

Abstract

Caused by vibration during the launch and changes of environmental factors, the camera is defocused by sensor focal plane shift. The defocus blur of images seriously hinders the application of remote sensing images (RSIs), so the accurate on-orbit refocusing of optical sensors has become a challenge for space agencies. By analyzing the blur kernel size of on-orbit thermal infrared images, a new defocus assessment method is proposed in this paper. First, the blur kernels of real-time RSIs are estimated by four state-of-art blind image deblurring methods. Then, to verify the reliability of the proposed method, the trend of the full width at half maximum of blur kernel and that obtained by traditional knife edge method are verified to be consistent. And the three-dimensional volume is also calculated as an appended evaluation. Compared with no-reference image quality assessment, the change of blur kernel sizes is proved to be a better evaluation of the sharpness for refocusing guidance. Specifically, smaller blur kernels imply sharper images. Different from the traditional defocus assessment method, the proposed method does not depend on the slanted-edge area and has better applicability in the field of remote sensing.