JPEG 2000 Extensions for Scalable Coding of Discontinuous Media.

Area of Science:

• Image and Video Compression
• Digital Signal Processing
• Computer Vision

Background:

• JPEG 2000 is a standard for image compression.
• Discontinuous media, such as depth maps and optical flows, present challenges for standard compression techniques.
• Existing methods may not efficiently capture boundary geometry in piecewise smooth imagery.

Purpose of the Study:

• To extend the JPEG 2000 standard for improved coding of discontinuous media.
• To develop a compression method that accurately models discontinuity boundaries.
• To maintain the scalability and accessibility features of JPEG 2000.

Main Methods:

• Proposed novel extensions to the JPEG 2000 framework.
• Utilized breakpoints to model discontinuity boundary geometry.
• Applied a breakpoint-dependent Discrete Wavelet Transform (BP-DWT) to input imagery.
• Encoded breakpoint and transform components as independent, progressively decodable bit streams.

Main Results:

• Demonstrated advantages of breakpoint representations with BP-DWT and embedded bit-plane coding.
• Achieved competitive rate-distortion performance compared to existing methods.
• Preserved the highly scalable and accessible coding features of JPEG 2000.

Conclusions:

• The proposed extensions effectively enhance JPEG 2000 for discontinuous media.
• Breakpoint modeling and BP-DWT offer significant improvements for depth maps and optical flows.
• These extensions are being standardized as Part 17 of the JPEG 2000 family.