Study of compression statistics and prediction of rate-distortion curves for video texture

Encoding textural content remains a challenge for current standardised video codecs. It is therefore beneficial to understand textures in terms of both their spatio-temporal characteristics and encoding statistics order optimise performance. In this paper, we analyse the features textures, explore rate-quality performance different texture types investigate models mathematically describe them. For all considered theoretical models, employ machine-learning regression predict curves based solely on selected extracted from uncompressed content. All experiments were performed homogeneous ensure validity observations. The results indicate that using an exponential model can more accurately expected curve (with mean Bjøntegaard Delta rate .46% over dataset), while maintaining low relative complexity. This be adopted by loop processes faster decisions such as rate–distortion optimisation, adaptive quantisation, partitioning, etc. • Study relation compression Video prediction low-level Comparative analysis computational