Unsupervised Vessel Segmentation Technique Using LTP and PCA

This paper presents a retinal vessels segmentation method than can be used by the ophthalmologists to diagnose different retinal diseases, such as thickening, thinning, or splitting. The proposed method consists of various interconnected steps. Initially, the G-channel is isolated from the input colored RGB image. Later, the Principal Component Analysis (PCA) is applied on the Green channel. Next, the Local Ternary Pattern (LTP) is used that feeds the processed image to the Morphological Filter Bank (MFB). The MFB uses various filters, such as spur, majority, and opening operations to segment the vessels image. Simulations are performed on the DRIVE and the STARE datasets. The proposed method yields mean accuracy of 94.07% and 93.45% on DRIVE and STARE datasets, respectively. Moreover, the proposed method is computationally less complex as compared to the few published methods.