Document fraud detection by ink analysis using texture features and histogram matching

This paper proposes an efficient method to detect any fraud document. It considers texture features, such as Local binary pattern and Gabor filters and performs a histogram matching to analyze the document. Texture features and RGB color information of each word in the document are extracted. Normalized histograms of two different images of a document are compared to generate a matching score for decision making. The advantage of the proposed method is that it is fully unsupervised and hence, it does not require any prior knowledge and is less time consuming. The method is found to be very efficient.