Evaluation of IQM´s effectiveness for cell phone identification using captured videos and images

Images and videos captured by cell phones are very important carriers of information amongst user generated multimedia contents. These carriers can be used in solving many forensic problems such as identification of movie piracy, insurance cases, child pornography, and other applications involving identifying/verifying source cell phones. This paper evaluates the effectiveness of Image Quality Measures (IQM) for identifying the source cell phone from the images or videos captured by that cell phone, by comparing the classification accuracies obtained for these two scenarios. Twenty-eight IQM features for each image and selected video frame are extracted and then classified using the Rotation forest classifier of WEKA. The proposed method is tested on 900 images and 1,350 short videos from nine different cellphones, some of which are of same brands and models. The experiments demonstrate that due to larger compression artifacts in videos, IQM are less effective for video based source cell phone identification as compared to image based source cell phone identification.