Abstract:

CFA (color filter array) interpolation will introduce correlations between different color channels. In order to improve the accuracy of source camera identification, a novel source camera classification method based on  inter-channel correlations of images was proposed. Variance maps of the green-to-red and green-to-blue spectrum differences were employed to reflect the differences of interpolation algorithms. Such differences could be used as the evidence to distinguish images taken by cameras of different brands or models. The proposed variance maps could efficiently and accurately differentiate interpolation algorithms. Experimental results showed that the proposed method can achieve high identification accuracy and good robustness for images captured by commercial cameras.

Key words: camera classification, image source identification, interpolation algorithm, color channels, variance map