Kmeans Clustering Based Ink Mismatch Detection

Forgery investigation and detection has been a relevant topic of interest for human beings since ages. Important messages written and transported by kings in old ages were sealed with signatures and stamps to achieve this purpose. But with the advent of digital technology, forgery detection has become even more important since tools for forgery have become vast as well. In this paper a technique based on pixel clustering has been introduced for detection of modification, alteration or forgery done with a different ink color pen. Hyperspectral images are used for ink mismatch detection in a handwritten note. We propose ink classification based on pixel intensities values present in all the bands of hyperspectral images of the handwritten note. Our proposed technique is quite simple yet effective in detecting ink mismatch with relatively high accuracy.