Performance of amplitude modified second-order Gaussian filter for the detection of retinal blood vessel

The profiles of blood vessel in color fundus image usually have no ideal edges. In this research, the matched filter is used to detect blood vessel instead of edge detector. Although existing matched filters are able to locate the vessels, they cannot easily measure the vessel diameters. An amplitude modified second order Gaussian filter is presented for the detection and quantification of the blood vessel. Mathematic analysis and simulation reveal that the vessel diameter can be measured by finding the maximum convolution peak. A vessel diameter measurement concept based on calibration is presented in order to avoid the burden of finding the exact edges of vessels. The real vessel width can be determined simply according to the relationship built through a calibration procedure in. An experiment result is given.