Automated surface texture classification of photographic print media

Digital imaging and signal processing technologies offer new, quantitative methods to classify and characterize surface textures. This paper presents a collaborative project to systematically and semi-automatically characterize the surface texture of traditional black and white photographic paper as well as contemporary photographic inkjet printing media. Surface texture is a critical feature in the manufacture, marketing and use of photographic papers, especially those used for fine art printing. Raking light reveals texture through a stark rendering of highlights and shadows. Though raking light images effectively document surface features of paper, the sheer number and diversity of textures prohibits efficient visual classification. This work provides evidence that automatic, computer-based classification of texture documented with raking light is feasible by demonstrating an encouraging degree of success sorting two sets of 120 close-up images made from diverse samples of inkjet paper and canvas available in the market from 2000 through 2011 as well as historic samples of black and white paper made at different times during the 20th century The samples used for this study were drawn from two prominent reference collections of photographic media: the Wilhelm Analog and Digital Color Print Materials Reference Collection and Paul Messier´s reference collection of historic black and white (silver gelatin) papers. Using these two datasets, four university teams applied different image processing strategies for automatic feature extraction and degree of similarity quantification. All four approaches were successful in detecting strong affinities among similarity groupings built into the datasets as well as identifying outliers. The creation and deployment of the algorithms was carried out by the teams without prior knowledge of the distributions of similarities and outliers. These results indicate that automatic classification of photographic paper based on texture - mages is feasible. To encourage the development of additional classification schemes, the two “training” datasets used in this work (comprising 240 images) is available to other academic researchers at www.PaperTextureID.org.).