A Customized Camera Imaging Pipeline for Dermatological Imaging
| Autor: | Iman Khodadad, Michael S. Brown, Farnoud Kazemzadeh, Hakki Can Karaimer |
|---|---|
| Rok vydání: | 2019 |
| Předmět: |
Color calibration
Computer science business.industry Machine vision ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION 02 engineering and technology 01 natural sciences Pipeline (software) Visualization 010309 optics Component (UML) 0103 physical sciences 0202 electrical engineering electronic engineering information engineering RGB color model 020201 artificial intelligence & image processing Computer vision Artificial intelligence business |
| Zdroj: | CVPR Workshops |
| DOI: | 10.1109/cvprw.2019.00329 |
| Popis: | This paper describes the customization of the camera processing pipeline of a machine vision camera that has been integrated into a hand-held dermatological imaging device. The device uses a combination of visible and non-visible spectral LEDs to allow capture of visible RGB imagery as well as selected non-visible wavelengths. Our customization involves two components. The first component is a color calibration procedure that ensures the captured images are colorimetrically more accurate than those obtained through the machine vision camera's native API. The need for color calibration is a critical component that is often overlooked or poorly understood by computer vision engineers. Our second component is a fast method to integrate the narrow band spectral images (some of which are outside the visible range) into the visible RGB image for enhanced visualization. This component of our pipeline involves evaluating several algorithms capable of multiple image fusion to determine the most suitable one for our application. Quantitative and subject results, including feedback from clinicians, demonstrate the effectiveness of our customization procedure. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|