Fast Depth and Mode Decision in Intra Prediction for Quality SHVC.

Area of Science:

• Video Compression Technology
• Digital Signal Processing

Background:

• Scalable High Efficiency Video Coding (SHVC) extends HEVC, offering high coding efficiency through complex intra prediction.
• The recursive quadtree-based division of Coding Units (CUs) and extensive mode checking (35 intra modes + Inter-Layer Reference) lead to significant computational complexity.

Purpose of the Study:

• To develop an efficient algorithm for accelerating intra prediction in SHVC enhancement layers.
• To reduce coding complexity and improve coding speed without sacrificing coding efficiency.

Main Methods:

• A Naive Bayes classifier combines temporal and spatial correlations to predict depth probabilities and skip unlikely depths.
• Inter-Layer Reference (ILR) mode probability and Partial Zero Blocks (PZBs) with Sum of Squared Differences (SSD) are used to efficiently determine ILR mode suitability, potentially skipping intra prediction.
• Initial Intra Modes (IMs) are derived using the Sobel operator and their Hadamard Cost (HC) values predict candidate IMs in Rough Mode Decision (RMD).
• An early termination criterion based on HC values is developed for Rate-Distortion Optimization (RDO).
• Depth selection is accelerated by combining depth probabilities and residual coefficient distributions.

Main Results:

• The proposed scheme significantly reduces the complexity of depth determination and mode decision.
• Experimental results show an average speedup gain exceeding 80%.
• Coding efficiency is maintained comparable to existing methods.

Conclusions:

• The developed algorithm effectively accelerates intra prediction in SHVC enhancement layers.
• It offers a substantial performance improvement by intelligently reducing computational complexity.
• The method provides a practical solution for faster SHVC encoding while preserving quality.