Research On Temporal Coded Exposure Imaging And Image Restoration

Motion image deblurring is an important problem in image processing.Based on mathematical model for motion image deblurring,the deblurring is represented by convolution of sharp image and kernel of imaging system,and the image is restored by inversion.However,the problem of inversion restoration is ill posed normally,and it is difficult to restore the moving target image accurately.In the traditional imaging mode,it is difficult for the image restoration to obtain the complete original image with the continuous opening the shutter.This is equivalent to the fact that there is a low-pass filtering process in the frequency domain,the high-frequency information is truncated and cannot be retained in the image.Coded exposure imaging technology divides the original continuous exposure time into several tiny chops,which are controlled by preset coding.Because the frequency of chops is higher than that of continuous exposure,the pass band of the frequency-domain filter is broadened in the imaging.When the amplitude of the preset coding in frequency-domain is far away from zero,more frequency information of the image is retained,especially the high-frequency detail information.Compared to continuous exposure mode,imaging technology of coded exposure improves the restoration of motion blurred image and provides a new idea for motion blurred image restoration.The restoration methods of single coded exposure image focus on the moving image with uni-directional relative motion,and the restoration must be assisted by peripheral measurement or according to the statistical law of natural image.In order to capture better coded exposure image for restoration,the principle of coded exposure imaging is studied systematically in this thesis.With the coded exposure imaging system,the method of coded exposure camera and image restoration with different movement is proposed.The main innovations are as follows:(1)Based on the principle of coded exposure,an embedded coded exposure imaging system is designed and built with CCD image sensor.Using CCD substrate control technology,photo-generated charge is readout only once by multiple intermittent charge accumulation according to the preset coding.In this scheme,the time of coded exposure image acquisition and charge driving is intensive highly,and the blur length of the blurred image is the same as ordinary camera.Meanwhile,blurring image is decoded to obtain a sharp restored image.The experimental results show that target's high-frequency information is protected and the problem of motion blurred image can be solved by the embedded coded exposure imaging system proposed and designed in this thesis.(2)Aiming at the restoration problem of the target with uni-directional motion,a joint estimation algorithm based on the similarity and entropy of the restored image is proposed.Structural similarity index is used to determine the blur length,based on the fact that similarity between the two structures of good restoration images and coded exposure images.By comparing the structural similarity index between the reconstructed decoded image and the coded exposure image,the search range is determined.In this range,the restoration image is found by the most ordered decode image.With the minimum entropy,the sharp restoration image is obtained and the automatic estimation of the blur length is realized.It is verified that the the kernel length and restored image can be estimated by the proposed algorithm with the simulation image and the actual image acquisition experiment.(3)Aiming at the limitation problem of the uni-directional motion mode in coded exposure,regularization method is used to solve the restoration problem with arbitrary motion.Compared to the traditional imaging mode,the continuous smear region of the blur image is segmented by coded exposure in the spatial domain,which makes the gradient strong relatively.In frequency domain,coded exposure keeps more image information,especially the high-frequency information of the image,which provides a guarantee for the subsequent restoration of high-quality image.This method not only retains the high frequency information in the coded exposure image,but also integrates the L0 regularization constraint formed by the gradient prior of the natural image in the image restoration.It is shown that the quality of the restored image obtained by adopted method is higher than that of the conventional image method in different motion conditions.