Inter-layer Prediction Using Enhanced Reference for Video Coding

The prediction process is segmented into two distinct paths: one for extracting high frequency information via high pass filtering, and another for extracting low frequency information via low pass filtering. This segmentation allows each path to optimize for its specific frequency range, resolving the contradiction by preserving both coding efficiency (through selective filtering) and prediction accuracy (through complementary frequency information combination).

The enhanced inter-layer reference picture is created by combining filtered versions of reference pictures, analogous to composite materials. The high pass filtered picture contributes high frequency components while the low pass filtered picture contributes low frequency components, creating a composite reference that maintains both the compression benefits and the prediction accuracy needed for enhancement layer coding.

Different quality processing is applied to different frequency components of the reference picture. High frequency components are preserved through high pass filtering while low frequency components are processed through low pass filtering. This local quality approach ensures that each frequency band receives appropriate processing, preserving high frequency information without uniformly increasing complexity across all picture data.

The filtering operations change key parameters of the reference picture data - specifically frequency domain characteristics. By applying high pass and low pass filters with specific cutoff frequencies and characteristics, the system transforms the reference picture into frequency-separated versions, preserving high frequency information while managing complexity through parameterized filter design.