Satomi Tanaka, Akihiro Kakinuma, Naohiro Kamijo, Hiroshi Takahashi, Norimichi Tsumura
Optical Review 2017 (accepted)
Abstract
The human visual system maintains the perception of colors of an object
across various light sources. Similarly, current digital cameras feature
an auto white balance function, which estimates the illuminant color and
corrects the color of a photograph as if the photograph was taken under
a certain light source. The main subject in a photograph is often a person’s
face, which could be used to estimate the illuminant color. However, such
estimation is adversely affected by differences in facial colors among
individuals. The present paper proposes an auto white balance algorithm
based on a pigmentation separation method that separates the human skin
color image into the components of melanin, hemoglobin and shading. Pigment
densities have a uniform property within the same race that can be calculated
from the components of melanin and hemoglobin in the face. We, thus, propose
a method that uses the subject’s facial color in an image and is unaffected
by individual differences in facial color among Japanese people.
[PDF] (to appeare)