Research On Image Restoration Method Based On Diffraction Spectrum Imaging Principle

Diffractive spectral imaging is a combination of diffraction techniques and the imaging properties of optical components.The diffractive optical imaging spectrometer(DOIS)has the advantages of high spectral resolution,high luminous flux,high signal-to-noise ratio,small volume,low price,and easy to realize lightweight and miniaturization.Therefore,it has a good application prospect in many fields such as atmospheric monitoring,image analysis,and vegetation classification.This dissertation focuses on the diffraction spectral imaging technology,mainly elaborates the imaging principle,system construction and image restoration.The specific work is as follows:1)This dissertation expounds the imaging characteristics and advantages of the diffractive lens,studies the basic imaging principle and image reconstruction process of DOIS,simulates the diffractive spectral imaging,and meanwhile builds a set of diffraction spectral imaging system to collect the experimental images.At the same time,the spectral curves of the blurred and the original sharp images are extracted.2)Study the blind deblurring method based on adaptive sparse prior,the blind deblurring method based on dark channel prior,and the Janson-Van Cittert(JVC)method for image restoration.These three popular image restoration algorithms were respectively used to recover the images captured by DOIS.By comparing the spectral curves,Laplace Gradient,Quality Index(QI)and Root-Mean-Square Error(RMSE),we found that the Janson-Van Cittert method has better performance for diffractive spectral images.3)Propose an improved inverse filtering algorithm with a regular filter based on regularization.The traditional restoration algorithms recover images only by kernel estimation or iteration without considering the imaging principle of diffractive spectral imaging.Therefore,the restored image loses the spectral information,so that we cannot extract the features of the ground.However,according to the imaging principle of DOIS,the improved inverse filtering algorithm can well separate the sharp image from the superimposed blurred images,and still preserve the spectral information while suppressing the noise to some extent by regularizing and fusing a regular filter.Through quantitative and qualitative comparative analysis,it can be seen that the improved algorithm proposed in this dissertation is much better than traditional methods in restoring image sharpness,fitting spectral curves and restoring spectral information,etc.Consequently,the improved inverse filtering algorithm can effectively improve the recovery performance for diffractive spectral images.