Research On Blood Vessel Centerline Extracting In Retina

Fundus blood vessel is an important basis for the diagnosis of disease, this article studied the anatomical characteristics of the eyeball, established the reconstruction parameters of suitable medical fundus visualization of mathematical model, converted the original color image type, enhanced the image, filtered image to get matched, thinned and tracked the inertial matrix to achieve the extraction of the retinal vascular centerline, in order to supply the three-dimensional parameters of the environment and the environment of the diagnosis fundus vascular angiography for clinical applications, providing useful help for laser treatment and surgical position.The paper is on the key technical problems in the retina for retinal image processing and analysis in research, the key technical problems includes retinal vascular network extraction, the width measurement of retinal vascular, retinal image registration and smooth stitching. Accurate extraction of retinal blood vessels in retinal image is the premise of analysis and processing in image, the segmentation method of blood vessel in retinal includes matched filter, region growth, morphology method, threshold segmentation and the edge gray level extraction method, statistical method, based on Hessian matrix method, based on tracing method and so on.Because of low gray level in retinal image, resolution between image and background is much too low, the range of blood vessel's widths ranges from one pixel to twelve pixels, especially the central intensity of some vessels from background is as few as four grey levels. It seems that something is wrong with detecting the retina boundary, optic disc and other pathology structure in extracting blood vessel centerline. In the paper the author make use of the threshold based on modified local entropy to segment the blood vessel in retinal and compute centerline from the processing image. The speed of modified method is improved, and small blood vessel branch can be excluded. The original entropy threshold primarily apply gray co-occurrence matrix from vertically lower direction and horizontally right direction to compute the intensity transitions. The modified entropy mainly make use of gray co-occurrence from the diagonal right direction to compute the intensity transition. The method is semi-automatic to trace the blood vessel centerline in retinal based on inertial matrix by appointing the initial point and using radius as scale and expanding the radius finding the optimum radius. The next tracing point is decided on the eigenvector of inertial matrix, iterative so until the blood vessel's centerline is tracked out.