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

Quantitative assessment of capabilities of colour models for pen ink discrimination in handwritten documents

“…Moreover, MLP‐based feature selection has also been used in [13] for identifying the adequate feature set. Gorai et al [15] have extracted local binary pattern and Gabor filter‐based ink features from the RGB colour model representation of the ink colours. Histograms of these features are created from each comparing word written using a particular ink.…”

“…At the same time, new words and numbers can also be added to manipulate the interpretation of existing words in a handwritten document. Approaches for ink‐based detection of forgery can be grouped into two categories, i.e., ink dating [3–8] and pen ink differentiation [9–22]. Ink dating deals with the forgeries in a handwritten document by analysing degradation and fading of the ink over a period of time.…”

“…Pen ink differentiation methods are further categorised into destructive [17–21] and non‐destructive [9–16, 22] methods. Traditionally, pen ink differentiation involves destructive techniques [17–21].…”

“…These techniques compromise the physical document itself, which is undesirable. The second category of pen ink differentiation techniques comprises of non‐destructive techniques, which includes image processing and pattern recognition based methodologies [9–16]. Due to this, the integrity of a document does not suffer loss, which is very important from the perspective of judicial standpoint.…”

“…It is observed that these techniques [9–16] vary in the colour model being considered in the representation of ink colour, the features representing ink sample, and the classifiers being used for discriminating the features from two different ink samples. Consider pair of words and written using two different pens as shown in Fig.…”