Optimization of retinal blood vessel segmentation based on Gabor filters and particle swarm optimization

Abstract

The structure of the retinal blood vessels can be obtained by segmenting the fundus images. A fundus image can be gained through color fundus photography or fluorescein angiography (FA). The fundus image produced by the camera can cause noise which can reduce the quality of the fundus image. To reduce the noise, this research uses the non-local means filter (NLMF). For texture analysis, the study uses Gabor filters due to the frequencies of this filter as the same as the human visual system. The segmenting process of the retinal blood vessel is performed using K-means optimized by particle swarm optimization (PSO). The accuracy of 0.9525, the precision of 0.8330, the sensitivity of 0.5817, and the specificity of 0.9880 are obtained using the proposed method.