Web conferencing and remote desktop applications require the desktop of one computer to be encoded and transmitted to be displayed on a remote screen. However, the basic profiles of common video codecs support chrominance sub-sampled format only, which may lead to annoying artifacts with screen content, as shown in the example.
The conversion process from full resolution to reduced-resolution chrominance channels consists of low-pass filtering followed by decimation, i.e., the actual sub-sampling. This process is typically performed in the spatial domain. However, this is also possible in the transform domain, e.g. after the block-based application of DCT-like transforms, as it is done in modern video codecs anyway.
Thus, the task of this Bachelor’s thesis is to convert between different chrominance resolutions in the transform domain. A simple conversion scheme has to be implemented, in order to analyze the effects of this conversion procedure and compare the results to traditional sub-sampling in the spatial domain. Moreover, also the potential of this approach to reconstruct the full chrominance in the transform domain shall be investigated. The focus of this work is on screen content data, which shall be used to evaluate the developed algorithms
The work includes a literature research. Implementation and evaluation are to be done in Python. The code has to be well-structured and documented. The thesis can be written in German or English.
Available from July 2018