A Novel Method for Enhancing Retinal Images using Vesselness

The motive of image enhancement is to focus on highlighting the hidden details of the image and removing the noise from the image. The research paper conducts the enhancement of the retinal images via CLAHE and four Morphological Operations (MOs). The Digital Retinal Images for Vessel Extraction (DRIVE) dataset of retinal images is used to conduct the research work. The extraction of the vessel profile spontaneously from the retinal images is a significant step in analyzing the retinal images. The proposed method conducted vesselness of retinal images at three different scales of 0.5, 1.0, and 1.5 for both colored and grayscale images. The proposed method makes use of Gaussian kernels to calculate Eigenvectors and eigenvalues, Histogram Equalization and Median filter to enhance the images, Gaussian filter to remove noise, and power law to sharpen the output image. The proposed method is more robust as compared to CLAHE and MOs. The enhancement achieved by the proposed methodology outperformed the enhancement achieved by CLAHE and morphological operations. The average PSNR of the proposed method is 57.39dB as compared to 36.11dB and 50.68dB for CLAHE and MOs respectively. The average value of MSE and RMSE for the proposed method is 0.3454 and 0.5872 as compared to 3.9919 and 1.997 for CLAHE and 0.7454 and 0.8633 for MOs respectively. The suggested approach can serve as an effective preprocessing tool for segmenting and classifying the DR-related feature methods. The authenticity of the work done is justified by calculating the values of PSNR (Peak Signal Noise Ratio), MSE (Mean Square Error), and RMSE (Root Mean Square Error) in each case which are used as performance evaluation metrics.