Research On The Application Of Retinex Methods In Different Low-quality Image Enhancement

As an effective image quality promotion tool,enhancement techniques are widely used in image processing.In recent years,Retinex theory has received a lot of attention from various fields for its color constancy,and Retinex-based enhancement algorithms are often used to improve the visual quality of images.In this thesis,we conduct various studies on the improved Retinex methods to address the problems of illumination and noise in low-quality vertebra Magnetic Resonance Imaging(MRI)images,retina Optical Coherence Tomography(OCT)images,and Electronic Speckle Pattern Interferometry(ESPI)fringe images.Besides,we explore their applications in segmentation and binarization.The main studies and innovations of this thesis are as follows:(1)In order to address the phenomenon of low and inhomogeneous contrast in MRI images,this thesis proposes an improved sequence decomposition Retinex enhancement based on the haze removal algorithm.The illuminance and reflectance are first estimated by the sequence decomposition Retinex model,and then a haze removal algorithm based on the dark channel priori is introduced to process the inverted illuminance.We provide qualitative and quantitative evaluation on low-quality vertebra MRI datasets,and compare the results with four advanced Retinex enhancement methods.The experiments demonstrate that the proposed method can improve the illumination and contrast of the images significantly and achieve a certain degree of noise suppression.(2)In order to address the problem of strong speckle and low structural contrast in OCT images,this thesis proposes a Selective Weighted Variational Enhancement(SWVE)model to enhance the structure parts of OCT images,and then present a ShapePreserving Fourth-Order Oriented PDE(SP-FOOPDE)algorithm to suppress speckle noise.Experiments on different kinds of OCT images show that the proposed method performs well in contrast enhancement and speckle suppression with good anti-noise ability and strikes a splendid balance between noise suppression and structure preservation.(3)In order to binarize the ESPI fringe patterns with massive inherent speckle noise and extremely low contrast,this thesis proposes a binarization method based on the multi-scale Retinex(MSR)enhancement and entropy.The method first utilizes MSR for enhancement,then calculates the local entropy and uses the Second-Order Oriented PDE(SOOPDE)algorithm for filtering,finally performs the Fuzzy C-Means(FCM)clustering on the filtering results of local entropy to achieve effective binarization.In this thesis,we conduct experiments on low-quality ESPI experimental subtraction fringe patterns and computer-simulated addition fringe patterns,and compare the results with the related methods.The results show that the proposed method can obtain satisfactory binarization and extract ideal skeleton line.