Image Processing Lab (IPL)
Camera identification in the presence of computational photography
Reference persons TIZIANO BIANCHI
Research Groups Image Processing Lab (IPL)
Thesis type RESEARCH / EXPERIMENTAL
Description Every camera sensor leaves imperceptible traces on the acquired images. One of these traces, due to slight differences in sensitivity of sensor pixels and named Photo Response Non-Uniformity (PRNU), is a well established technique for linking an image to the sensor that acquired it. Unfortunately, in modern smartphones those traces can be significantly altered by the software pipeline leading to image formation, known as computational photography. Sophisticated algorithms are routinely used to insert customized filters or apply HDR effects. The effect is that the standard PRNU test can give a large number of false positives, since PRNU traces of different sensors tend to look more similar after they are processed by the same computational photography algorithm.
The aim of this thesis is to develop an innovative test for camera identification that is not affected by computational photography, possibly exploiting recent advances on PRNU estimation based on machine learning.
Iuliani, Massimo et al. “A leak in PRNU based source identification? Questioning fingerprint uniqueness.” https://arxiv.org/abs/2009.04878
Required skills Signal processing skills, Matlab and/or Python programming. Background on neural networks is a plus.
Deadline 05/10/2022 PROPONI LA TUA CANDIDATURA