A fourier transform- and training-based method for recognizing arbitrary color filter array structure

Assume that an input mosaic image with arbitrary CFA structure has lost its header, this paper presents a novel efficient method for recognizing its CFA structure using the frequency domain approach. The proposed method consists of a training-based scheme and an identification scheme. Initially, based on a set of training mosaic images with different CFA structures, a training-based scheme is proposed to build up a model map for every CFA structure. Firstly, the proposed three-step identification scheme performs the same steps used in the construction of model maps for the header-less input mosaic image to obtain the query map. Then, a matching scheme is proposed to identify the corresponding CFA structure of the query map from the model maps based on the distance criterion.