On the Influence of Texture Strength and Camera Signal on Micro-Texture Surface Classification

The digital cameras contained in nearly all modern smartphone devices can be utilised for a broad range of applications, e.g. biometric recognition and extrinsic product authentication which are well-established and work reliably. However, previous research revealed undeniable problems with micro-structure surface classification if different smartphone devices are involved (cross-device scenario). This work is a step further towards understanding the underlying reasons by employing different training strategies (training on general texture patches and testing on the dental ceramic ones), analysing the texture properties (information entropy) and testing the performance of pairs and triples of the same make and model of smartphones compared to arbitrary device pairs.