Fundus Image Restoration Based On Adaptive Blind Deconvolution

Fundus image which captured by the fundus camera is a standard objective diagnostic image. Retinal microvascular is the only deeper capillaries that can observed non-invasive dir ectly, which plays a vital role in the early discovery diagnosis and treatment of the diabetes, hi gh blood pressure, cerebral arteriosclerosis, coronary atherosclerosis and other cardiovascular cerebrovascular diseases. Getting a clear picture of retinal blood vessels is a prerequisite for a ccurate diagnosis of the disease. So the study of this subject has great significance on research ing body disease, early diagnosis by the non-invasive retinal image.In the fundus image acquisition process, there are several reasons inevitably lead to ima ge degradation. Fundus degradation image has local uneven illumination, low contrast and lo w noise ratio SNR and other features. The blurred image brings varying degrees of inconvenie nee for observation and diagnostic. The restoration algorithm of this paper, can improve the i mpact of local uneven illumination, and get the clear fundus image.This thesis studies the structure and characteristics of the human eye and fundus camera’s lighting system, imaging system, also analyzes their impact on the imaging process and the causes of retinal image blur.Then compare the color image processing component selection. Analysis and comparison of RGB and HSI color space, draw HSI color space I channel signal to noise ratio for color fundus image restoration. Therefore, this article puts color image restoration, into I channel restoration of HSI color space, and then convert the image to RGB color image.After getting the analysis and comparison of a variety of image restoration methods, for the case of a lack of prior knowledge of the fundus image, the blind deconvolution algorithm is selected. After the establishment of image restoration model in estimating the degraded image point spread function at the same time, the use of adaptive control algorithm iterations, take blind deconvolution algorithm restored image. After recovery, color and gray fundus image clarity, increased by 16.4% and 15.8%. And the restored image deconvolution algorithm comparison with the other. This article also adopt smooth boundary gradient algorithm to suppress "ring" effect of blind restoration image. After this, we analysis the effect of image restoration by the parameters of image sharpness.Finally, according to the actual demand, fundus image restoration system is developed on the system software platform. This system can achieve effective rehabilitation fundus images and provide clear fundus image for the diagnosis. It also provides a reference for the other image processing technology and diagnosis analysis.