Directional filtering transform for image/intra-frame compression.

Area of Science:

• Digital image processing
• Video compression algorithms
• Signal transforms

Background:

• Traditional transforms in image/video compression can yield different results based on the order of 1-D transforms.
• Anisotropic image models are used to analyze the impact of transform order on coding gain.
• Practical compression schemes often simplify decomposition, leading to performance variations with transform order.

Purpose of the Study:

• To analyze the theoretical coding gain affected by transform orders in 2-D transforms.
• To propose an adaptive transform order solution for practical compression scenarios.
• To introduce a novel directional filtering transform (dFT) for enhanced H.264 intraframe coding.

Main Methods:

• Theoretical analysis of transform orders using an anisotropic image model.
• Development of a directional filtering transform (dFT) with adaptive transform order.
• Integration of dFT into H.264 intraframe coding to exploit interblock and intrablock correlations.
• Experimental evaluation of the proposed scheme in H.264 intraframe coding.

Main Results:

• Transform order has minimal effect on coding gain with full decomposition and good directional modes.
• In practical schemes, transform order significantly impacts coding performance due to incomplete decomposition.
• The proposed dFT with adaptive order demonstrates superior objective and subjective performance in H.264 intraframe coding.

Conclusions:

• Adaptive transform order is crucial for optimizing practical compression schemes.
• The proposed directional filtering transform (dFT) effectively exploits image correlations for improved H.264 intraframe coding.
• The dFT offers an evenly distributed set of prediction modes and adaptive transform ordering for enhanced video compression.