Native Resolution Detection of Video Sequences
Ioannis X. Katsavounidis, Anne Aaron, David Ronca · 2015
We present a method to detect the lowest resolution that a video sequence segment has undergone before being rendered and presented at some higher resolution. The method is based on extracting homogenous regions within a given video sequence, performing Discrete Fourier Transform on each one of the color components (Y/Cb/Cr) for each video frame, normalizing the magnitude of the spectrum such that the total AC energy is one and then taking its logarithm. Average log-magnitude spectra demonstrate certain patterns that are typical of re-sampling techniques that video frames are undergone during post-production. The method is demonstrated on a few video sequences, showing typical up-sampling of video content, as a result of inherent camera resolution limitations, special effects overlaying at lower than the natural video scene footage and chroma subsampling.