Color constancy in dermatoscopy with smartphone
View/ Open
Author
Cugmas, Blaž
Pernuš, Franjo
Likar, Boštjan
Date
2017-12-07Metadata
Show full item recordAbstract
The recent spread of cheap dermatoscopes for smartphones can empower patients to acquire images of skin lesions on
their own and send them to dermatologists. Since images are acquired by different smartphone cameras under unique
illumination conditions, the variability in colors is expected. Therefore, the mobile dermatoscopic systems should be
calibrated in order to ensure the color constancy in skin images. In this study, we have tested a dermatoscope
DermLite DL1 basic, attached to Samsung Galaxy S4 smartphone. Under the controlled conditions, jpeg images of
standard color patches were acquired and a model between an unknown device-dependent RGB and a device independent
Lab color space has been built. Results showed that median and the best color error was 7.77 and 3.94,
respectively. Results are in the range of a human eye detection capability (color error ≈ 4) and video and printing
industry standards (color error is expected to be between 5 and 6). It can be concluded that a calibrated smartphone
dermatoscope can provide sufficient color constancy and can serve as an interesting opportunity to bring dermatologists
closer to the patients.