Retinal Image Enhancement Using Hybrid Approach

Verma, Prem Kumari; Singh, Nagendra

doi:10.1007/978-981-16-9650-3_40

Cited by 2 publications

(1 citation statement)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In order to streamline the network, Hajabdollahi et al [18] presented a CNN approach for segmenting retinal images by trimming convolutions layers and quantizing the fully connected layers. Sushil Kumar Saroj et al [19] proposed a model using a matched filter approach with probability distribution functions.In this paper consist the pre-processing model such as PCA and CLAHE technique [20] for the enhancement of retinal blood vessels. after this approach authors apply the Frechet Match filter for the segmentation process and found the accuracy with 95.44%.…”

Section: Related Workmentioning

confidence: 99%

Patching based Deep Learning approach for Retinal blood vessels segmentation

Garg

Verma

Singh

2023

Preprint

Self Cite

View full text Add to dashboard Cite

Human Eyes are one of the important organs of the humanbody.They are also very prone to diseases such as cataracts, AgeRelated Macular Degeneration, Diabetic Retinopathy, Glaucoma, etc.Therefore, it becomes important to develop methods that can helpin the diagnosis and treatment of these diseases before it is toolate. In this paper, a time-efficient and accurate deep learningbased bio-medical image segmentation approach is presented whichcan lay the foundation for further development in this area.The u-net deep learning model used in the proposed approach, was specially developed to deal with medical images. The proposed approachhas been split into three notable phases: data pre-processing, training U-net deep learning model, and data post-processing. The U-netmodel is fed this data and trained until satisfactory results. Duringpost-processing, corresponding predicted images are merged to obtain toprediction for the original images. The final step includes the removalof padding so that the original dimensions of images are maintained After a long period of experimentation, quantitative performance metrics have been respectively calculated as 0.963, 0.992,0.910 for the DRIVE data set and 0.958,0.993,0.888 for theSTARE data set respectively. Values of Loss have been 0.106,and 0.106 for the DRIVE and STARE data sets respectively.

show abstract

Section: Related Workmentioning

confidence: 99%