The colors of a scene are strongly dependent on the light source. For this reason, the systems for image retrieval or object recognition based on the analysis of descriptors related to the colors, cannot work well when illuminant changes occur. This work proposes a novel method for estimating the changes of illuminants occurring between the images of a same scene captured under different photometric conditions. The illuminant variations are represented by the von Kries diagonal model and their estimate is performed by minimizing a dissimilarity measure between the color histograms of the images under examination. A technique for illuminant invariant image and object identification based on the estimates of the light changes is also presented, along with a comparison with other approaches.
Estimating Illuminant Changes in Color Images by Color Histogram Comparison
Lecca, Michela;Messelodi, Stefano
2008-01-01
Abstract
The colors of a scene are strongly dependent on the light source. For this reason, the systems for image retrieval or object recognition based on the analysis of descriptors related to the colors, cannot work well when illuminant changes occur. This work proposes a novel method for estimating the changes of illuminants occurring between the images of a same scene captured under different photometric conditions. The illuminant variations are represented by the von Kries diagonal model and their estimate is performed by minimizing a dissimilarity measure between the color histograms of the images under examination. A technique for illuminant invariant image and object identification based on the estimates of the light changes is also presented, along with a comparison with other approaches.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
