Research On Restoration And Enhancement Of Postmortem Individual Eye Fundus Images Based On Dark Channel Prior And Image Information Entropy

Objective: Retina-based identity recognition is a current recognition scheme with strong stability,accuracy and application significance.However,due to disease or death of individuals or tissues,the cornea has become turbid,causing the collected fundus images to be cloudy and unclear.The main purpose is to design a set of fundus image defogging and enhancement algorithms to restore and enhance the original information characteristics of the fundus image to facilitate the subsequent application of the fundus image.Methods: According to the atmospheric scattering model,the algorithm in this paper first uses linear contrast enhancement to increase the difference between the spot and the target area in the image,selects brightness and gradient thresholds to segment the image into the target area and interference area,and selects a certain range based on the color attenuation prior.Iterate on the saturation and brightness characteristics of the image,use the iterative best transmittance image and atmospheric light component to deblur the original image,and then use the information entropy feature to iteratively filter out the appropriate overlay weights,and realize the self-control of the image blood vessels.Adaptation enhancement.In this paper,30 fundus images in the DRIVE database are selected for simulation and fogging processing,and 20 adult white rabbits are selected to shoot the fundus image set after death.The above images are processed by the algorithm in this paper,and the similarity of image structure is used to verify and evaluate the algorithm.Results: The simulated fog image set processed by the algorithm in this paper,the structure similarity SSIM distribution is roughly a normal distribution with 0.6166 as the mean and 0.1113 as the standard deviation.The total average time is about 1.490 s,which is similar to the average processing cost of the Fattal algorithm,the time is similar to 1.227 s.For the rabbit eye fundus images divided into 6 groups according to the time of death,the mean image information entropy is 7.72±0.21?7.47±0.24?7.07±0.52?7±0.21?6.34±0.21?6.32±0.34;after processing by this algorithm;the mean PSNR is 21.71±2.82?20.15±2.52?17.52±2.96?17.01±2.53?15.21±1.25?15.9±1.73,the total average time is about 2.143 s,which is similar to the average processing time of Fattal algorithm,that time is about 1.812 s.Conclusion: The algorithm in this paper has a strong ability to recover fundus images of corneal opacity.After processing by this algorithm,the recovery of fundus image information is more significant,and it has good adaptive ability in the processing process.It can be more effective for images with different degrees of blur.The fundus blood vessel structure is well reproduced,and the image details can be adaptively enhanced.In the restored fundus image,the large and small blood vessels,optic discs and bifurcation points have been clearly enhanced.