Accurate vessel segmentation using maximum entropy incorporating line detection and phase-preserving denoising

The retinal images with lesions, exudates, non-uniformed illuminations and pathological artifacts have intrinsic problems such as the absence of thin vessels and false vessels detection. To solve these problems, we propose a novel algorithm which involves separation of background images to minimize the influence of noise, non-uniformed illuminations and lesions. We develop two different strategies to segment thin and thick blood vessels. Thin blood vessels are identified by taking benefits of local phase-preserving denoising, line detection, local normalization and maximum entropy thresholding. To remove noise and preserve detailed blood vessels information, phase-preserving denoising technique is used. The technology takes an advantage of log-Gabor wavelet responses in the complex domain to preserve the phase information of the image. Thick vessels are extracted and binarized via maximum entropy thresholding. The performance of the proposed algorithm is tested on four popular databases (DRIVE, STARE, CHASE_ DB1, HRF). The results demonstrate that the proposed segmentation process is automatic, accurate and computationally efficient.