Color Cast Detection Method Based on Multi-feature Extraction - Advances in Image and Graphics Technologies - page 104

Methods based on prior knowledge [5] are examples of the latter group. Such me-
thods need to store prior knowledge to helpclassification.Similarapproaches include
methods based on machine learning [6].But the existing methodscannot tell the color
cast images apart exactly.
Therefore, in this paper, we choose AdaBoost to do the training and classification.
And based on the theory of color constancy, we bring up with a color cast detection
method based on multi-feature extraction.
The paper is organized as follows:In section 2, the method based on AdaBoost and
feature extraction is discussed. In section 3 and section 4, the experimentsare carried
out on the Ciurea database. Finally, in section 5, the conclusion is provided.

2 Color Cast Detection Method Based on AdaBoost

Mankind has the ability to correct the color cast in the scene adaptively, which
causes the object to be perceived constant along with the changing illuminant.And
the ability is called the theory of color constancy. Taking this knowledge into con-
sideration, we found that the color cast is not only related to the average and the
variance of the chromaticity information, but also linked to the distribution of the
chromaticity information [4]. In this paper, we will extract four kinds of features
and then use the AdaBoost algorithm to train and classify. Fig.1 shows the
workflow of the method.

Fig. 1. Workflow of the method