0x09/resdet

What it solves

resdet is designed to detect if an image or video has been upscaled and, if so, identify the original resolution it was scaled from. This is particularly useful for identifying the source quality of media that has been enlarged using traditional resampling methods.

How it works

The tool uses the Discrete Cosine Transform (DCT) to analyze the frequency domain of the image. Traditional resampling methods leave specific signatures—zero-crossings in the frequency extension—that occur at offsets corresponding to the original resolution. resdet identifies these inversions to guess the source dimensions.

Who it’s for

It is intended for users who need to verify the original resolution of media, such as those working in image deduplication or digital forensics, and developers who want to integrate this functionality via a C library with Python and JavaScript bindings.

Highlights

• Multi-format support: Works with images and videos, utilizing loaders like FFmpeg, libpng, libjpeg, and MagickWand.
• Cross-platform library: Written in standard C (C99) and can be used as a standalone library (libresdet).
• Language bindings: Provides bindings for Python and JavaScript (via Emscripten).
• Frequency analysis: Leverages FFT implementations like KISS FFT or FFTW for high-performance spectral analysis.