Compensation with Multi Motion Models for Images from TDI-CCD Aerial Camera

The paper presents the method to restore images in a blind way from TDI -CCD camera in aerial environment, based on some motion models. Following the shooting environment and principle of TDI-CCD imaging, three models are chosen for image blurring: from forward image motion, from random vibration in low frequency because of atmospheric turbulence, from aircraft vibration in high frequency. Before blind restoration begins, the value ranges for parameters existing in the three models are determined utilizing apriori information from the aerial shooting environment and conditions. And then, the best image results restoring from separated variable parameters are judged, following the trend of edge gradient index changes in terms of the variational parameters with the aid of visual inspection. The parameters include: offset value of forward motion, σ value for random vibration, vibration frequency, amplitude and exposure time for vibration in high frequency. Those corresponding to the best image results are the optimum ones. After all optimum parameters set into the motion models, image can be restored using Wiener filtering. The images from TDI-CCD camera get good restoration results with the blind restoration method addressed in the report, which means the three motion models can really represent the mechanism of image blurring. Also, the method using edge gradient index and the apriori information in aerial shooting environment can judge the parameters in the models correctly and half-automatically.