SEMI-AUTIOMATIC VIDEO INDIRECT OPHTHALMOSCOPY PHOTOMONTAGE

Permalink

https://hdl.handle.net/2142/124796

Title: SEMI-AUTIOMATIC VIDEO INDIRECT OPHTHALMOSCOPY PHOTOMONTAGE
Author(s): Wei, Jionghao
Issue Date: 2023-05-01
Keyword(s): Image Processing; Video processing, Computer Vision; Image Registration; Machine Learning
Abstract: Indirect Ophthalmoscopy is a retinal examination in which ophthalmologists use a strong light and specific lens to investigate the back of the eyes. It is proven to be valuable for diagnosis and treatment of retinal tears, holes, and detachments by providing a wider field of view. Recent Indirect Ophthalmoscopy headsets acquire recording ability so that ophthalmologists can share the recordings to communicate. However, due to the nature of the examination, the recordings contain many blurred and irrelevant frames. Furthermore, reviewing the images taken by an indirect ophthalmoscope is cumbersome as images are acquired from multiple views and contain overlapping information. To facilitate the review of indirect ophthalmoscopy and communication between ophthalmologists, we propose a semi-automated, machine-learning pipeline to filter frames based on retinal features and blurriness, and perform photomontage using the clearest frames.
Type of Resource: text
Language: eng

The best of ECE undergraduate research