Improving PRNU Compression Through Preprocessing, Quantization, and Coding

摘要: In the last decade, extremely rapid proliferation of digital devices capable acquiring and sharing images over Web has significantly increased amount publicly accessible by everyone with Internet access. Despite obvious benefits such technological improvements, it is becoming mandatory to verify origin trustfulness shared pictures. Photo response non-uniformity (PRNU) reference signal for forensic investigators when comes verifying or identifying which camera device shot a picture under analysis. spite this, PRNU almost white-shaped noise, thus being very difficult compress storage large scale search purposes, are frequent investigation scenarios. To overcome issue, community developed series compression algorithms. Lately, Gaussian random projections have proved achieve state-of-the-art performance. this paper, we propose two additional steps that help improving even more rate: 1) decimation preprocessing step tailored at attenuating frequency components in traces already suppressed JPEG compressed 2) dead-zone quantizer (rather than commonly used binary one) enables an entropy coding scheme save bitrate storing fingerprints sending residuals communication channel. Reported results show effectiveness proposed both controlled real case scenario.