Retinal vessel segmentation and delineation of morphological attributes of retinal blood vessels are utilized for diagnosis, screening and evaluation of various ophthalmologic diseases. In this paper, local entropy-based thresholding and modified Gaussian-based matched filter segmentation techniques have been performed on the bit plane sliced images to assess the information of blood vessels present in different bits. Bit plane slicing has been used prior to thresholding for efficient segmentation and further processing as it highlights the contribution made to the total image appearance by specific bits. This will help in efficient transmission of retinal data and diagnosis of diseases in retinal images. The efficiency of the segmentation method is calculated by evaluating performance measures. This is useful for image compression, robust person identification and efficient transmission. Local entropy-based thresholding is used on the reconstructed image, obtained by combining specific bits in bit plane slicing, for segmentation. This method is tested on the publicly available DRIVE and Aria databases. Performance measures are calculated for the segmentation methods. Supervised segmentation on bit planes sliced images performs better than Gaussian matched filter method.

Bit plane slicing local entropy-based segmentation matched filter specificity