Performance analysis of blood vessel in retinal image using simple approach detection / Mohd Rizman Sultan Mohd

Different image transformations through structuring element construction have been proposed to do an analysis of blood vessel by using retinal image. The Top-hat transformation technique had been widely used in this approach as for improving the previous method called entropy thresholding. This paper introduces modified Top-hat transformation by using Bottom-hat transformation to achieve better sensitivity and accuracy in transforming the fundus image to help detecting the blood vessel in retinal images. The step in this approach start of extracting the green channel of RGB image and it then will be enhanced by using the proposed method. The step is furthered by a mask generation and match filter followed by applying a local entropy thresholding and length filtering for vessel extraction. The performance of the result is determined by the sensitivity and accuracy calculation performed by comparing with existing hand-labelled results from the database. The difference between Top-hat and proposed techniques were then analysed. This project is done by using MATLAB R2012a with the set of retinal image obtained from STARE database. The proposed method manages to obtain Sensitivity of 0.85 and Accuracy of 0.95.