Exploiting Intra-Slice and Inter-Slice Redundancy for Learning-Based Lossless Volumetric Image Compression.

Area of Science:

• Computer Vision
• Image Processing
• Data Compression

Background:

• 3D volumetric image processing requires efficient storage and transmission due to large data volumes.
• Existing lossless compression techniques for volumetric data face challenges in handling complex spatial redundancies.

Purpose of the Study:

• To develop the first end-to-end optimized learning framework for lossless 3D volumetric data compression.
• To improve entropy modeling by integrating intra-slice and inter-slice information.
• To enhance the performance of lossless volumetric image compression.

Main Methods:

• Proposed an end-to-end learning framework for lossless 3D volumetric compression.
• Introduced hierarchical intra-slice auxiliary features extracted via multi-scale modules.
• Developed an Intra-slice and Inter-slice Conditional Entropy Coding module with Intra-Gate and Inter-Gate mechanisms for optimal feature fusion and entropy estimation.

Main Results:

• The framework achieved superior lossless compression performance on volumetric data.
• Demonstrated significant improvements over state-of-the-art hand-crafted and learning-based codecs.
• Validated effectiveness on 3D Medical Images and Hyper-Spectral Images datasets.

Conclusions:

• The proposed end-to-end framework offers a significant advancement in lossless 3D volumetric image compression.
• Jointly learning intra-slice and inter-slice features enables more accurate entropy estimation.
• The method shows strong potential for practical applications in medical imaging and remote sensing.