A Robust Segmentation of Blood Vessels in Retinal Images
| Autor: | Sheryar Mehmood Awan, Farhana Siddique, Talha Iqbal, Zahid Mahmood, Gul Zameen Khan |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Computer science
business.industry Matched filter media_common.quotation_subject 05 social sciences 0507 social and economic geography Retinal chemistry.chemical_compound chemistry Feature (computer vision) Contrast (vision) Segmentation Computer vision Artificial intelligence Noise (video) business 050703 geography Opening Communication channel media_common |
| Zdroj: | FIT |
| DOI: | 10.1109/fit47737.2019.00025 |
| Popis: | Retinal blood vessels generally appear as wire mesh structures that have different widths. Their morphology plays a crucial role in indication and timely handling of fatal diseases, for instance diabetic retinopathy or hypertension. To address the aforementioned issues, this paper presents a robust vessel segmentation technique. First, the input retinal image is splitted into the Red, Green, and Blue Channels. On the Green channel pre-processing is done to eradicate noise or other disease. This step also extracts key feature, such as bright or dark contrast of the input retinal image through top-hat morphological opening operation. Meanwhile, the Gaussian matched filters are applied along with different scales to achieve segmentation of thick or thin vessels. Finally, the pixel-by-pixel OR operation is applied to obtain final segmented image. Simulations on DRIVE and STARE datasets indicate that proposed technique surpasses recent methods in terms of Accuracy (Acc). The proposed technique is also computationally efficient than most of the recent methods compared therein. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|