SEGMENTASI PEMBULUH DARAH PADA CITRA RETINA DENGAN METODE NEIGHBORHOOD AVERAGE MENGGUNAKAN BACKGROUND EXCLUSION DAN THRESHOLDING OTSU
Daftar Isi:
- Blood vessels are an important part of the retina of the eye compared to other parts. This study presents a method for blood vessel segmentation in retinal images. In the first step, convert from the original image to Grayscale, then add the Gamma Correction value, and run the CLAHE process aimed at improving the quality of the retinal image. To reduce noise in retinal images, the Neighborhood Average method is then performed using Background Exception. In addition, blood vessels are segmented with the help of the Otsu Thresholding method. In the final step, Remove Small Objects, Incomplement, Median Filter, and Morphology Closing are applied. The proposed method allows for easier implementation and less computing time. The dataset used in this study is to DRIVE and STARE. In this case, the average value obtained for the DRIVE dataset is 94.73% accuracy, 49.16% sensitivity, 99.39% specification, 89.60% precision, and F1 score 63.27%. The STARE dataset with an accuracy of 93.11%, sensitivity 53.58%, specifications 97.41%, accuracy 70.88%, and F1 score 60.72% (compared to the Truth of Adam Hoover's Land), and Accuracy 91.93% , Sensitivity 47.53%, Specifications 98.15%, Precision 78.92%, and F1 Score 59.02% (compared to Valentina Kouznetsova's Ground Truth). Based on these results, this method can be proposed in the segmentation of blood vessels in retinal images.