Classifying the illumination condition from two light sources by color histogram assessment.

Abstract

We investigate and propose a method for assessment of the illumination condition covering two light sources. The method may be of some support for color vision and multispectral analysis methods that rely on a specific illumination condition. It is constrained to classifying the illumination condition for dielectric objects illuminated by two light sources. The reflected light is modeled by the dichromatic reflection model, which describes the light as the sum of its body reflections and surface reflections. Further, reflected light from an object illuminated by two light sources may give from one to four primary reflections depending on the condition, and it may be expressed as an additive mixture of these reflections. An additive mixture of two reflections expressed in chromaticities is limited to falling within the area enclosed by the chromaticities of the primary reflections of the light sources. So after finding the set of primary chromaticities enclosing the pixel points' chromaticities, it is possible for one to assess the current illumination condition. Since the method operates on pixel points globally, it is independent of illumination geometry and hence may be used on irregular objects. Two experiments are performed. One uses regular objects in a well-controlled laboratory environment and demonstrates that the pixel-point distribution is as expected. The second experiment demonstrates the method's potential use in support of spectroscopic analysis of vegetation through assessing the illumination condition of barley plants in an outdoor illumination condition.