Identification of Printing Paper Based on Texture Using Gabor Filters and Local Binary Patterns

There are many causes of deformation in an image and one of which during its acquisition to a digital image. The deformation takes different forms or causes different effects on the acquired image comparing with the original image including poor resolution, shear, noise, variation in the intensity and etc. A paper scanned by a scanner is a good example of possible deformation in images. Consequently, paper texture identification or fingerprinting is one of the research fields of pattern recognition that exposed to the deformation problem. Applications such as documents authentication deemed to be constrained by the deformation problem. Subsequently, one of the well-known methods in images texture extraction is the Locale Binary Pattern (LBP) method. However, the LBP method show a number of drawbacks in paper images texture extraction and two of which are neglecting some texture information of the images and incompetent to some images deformation due to its local view. In this paper combinations of Gabor filters and a LBP operator are proposed to reduce the effects of the mentioned drawbacks in papers fingerprinting domain. We use self-collected textures from 102 paper images in the test. The images are acquired in three resolutions of 50 DPI, 100 DPI and 150 DPI in order to manifest robust results. Consequently, the testing results of the proposed combinations improve paper images identification accuracy. This paper finds that applying Gabor filters prior to LBP method improve the LBP operator description and the fingerprinting accuracy.