Low Quality Retinal Blood Vessel Image Boosting Using Fuzzified Clustering

Retinal imaging can effectively diagnose diseases that manifest changes in the retinal anatomy. However, manual diagnosis paradigms are both error-prone and cost-intensive. Therefore, computer-aided technologies were developed for an exhaustive examination of retinal pathology and anatomy. In this a...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:IEEE transactions on artificial intelligence Ročník 5; číslo 6; s. 3022 - 3033
Hlavní autoři: Sinha, Sanya, Bhandari, Ashish Kumar, Kumar, Reman
Médium: Journal Article
Jazyk:angličtina
Vydáno: IEEE 01.06.2024
Témata:
Brightness
Brightness enhancement
contrast enhancement
Diseases
fuzzy c-means clustering
Histograms
Image color analysis
Image segmentation
Imaging
Retina
retinal images
ISSN:2691-4581, 2691-4581
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:Retinal imaging can effectively diagnose diseases that manifest changes in the retinal anatomy. However, manual diagnosis paradigms are both error-prone and cost-intensive. Therefore, computer-aided technologies were developed for an exhaustive examination of retinal pathology and anatomy. In this article, a new retinal image enhancement method based on fuzzy c-means is proposed to enhance low quality retinal blood vessel images while preserving its brightness. Fuzzy c-means clustering groups the intensity levels into multiple clusters and assigns a cluster membership value to each intensity level. These values are subsequently modified and are then mapped to their corresponding initial values. The green channel of a modified image obtained above is equalized using the adaptive histogram equalization to yield the enhanced image. The results for the proposed algorithm were established using standard datasets consisting of 1000 fundus images with 39 categories. The proposed technique preserves the brightness and improves the contrast while improving vascular segmentation.
ISSN:2691-4581
2691-4581
DOI:10.1109/TAI.2023.3336612