"Forensic analysis of print using digital image analysis"

61-72
The paper gives details of how small imperceptible differences in printed fonts using laser printers can be quantified using digital image analysis. The significance of the work is that if imperceptible differences in print can be measured then the techniques that can do this could have applications in forensic document analysis and security printing. The research required the development of an algorithm that can quantify features in perceptibly very similar text characters from noisy digital image data.

With digital technology becoming cheaper and less specialised, the techniques investigated here may become integrated into computationally intelligent mobile phone devices. Such devices could be used to check documents quickly and efficiently.

The key to the technique was the production of a skeleton of text characters from their digitally captured image. A skeleton is a line defined by a binary (0,1) value. The location of the binary line defines the edges of the text character from which accurate length measurements of the text character can be automatically taken. An algorithm processes the data to produce a shape factor. The computer then recognises a text character with a particular font by comparing its shape factor with the shape factors from different text characters in its database. The algorithm was developed solely by the author to carry out this task. The results of the research were that small differences in fonts could be detected by high resolution digital image analysis, and be used to objectively quantify small differences in font style.

Conference papers were selected and subjected to review by the editor and the conference program committee. They comprise electronic imaging and cognitive research specialists from international universities, and commercial organisations.