loading page

Ink Mismatch Detection in Hyperpectral Image using K -- means Clustering Algorithm
  • Usman Zia
Usman Zia
Institute of Space Technology

Corresponding Author:[email protected]

Author Profile

Abstract

In forensic analysis of any document, Ink mismatch detection provides a very important clue to see whether the document is corrupted or not. In hyperspectral imaging (HSI) ink mismatch detection is used to check whetherthe specific document is written with same pen/ink or different types of pens/inks have been used. In this paper, we show that aHSI of handwritten notes can differentiate between the ink sused to produce the document which are very similar in appearance. The HSI cube we have used for our research has 149 Bands. By using an unsupervised clustering algorithm, we have analyzed the spectral responses of the ink pixels to differentiate between the number of inks used in document. After the successful ink mismatch detection, we have used color labelling to classify the text written with different inks. This technique when applied to a RGB document, is not capable to differentiate between the number of inks used in the document. We hope that our findings and technique will help to examine HSI documents for forensic analysis.