Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
ISO 26429-4:2008 Digital Cinema Packaging — MXF JPEG 2000 Application
JPEG 2000 2K/4K profile mapping into MXF Generic Container for D-Cinema distribution
1. JPEG 2000 Mapping for D-Cinema MXF
ISO 26429-4 specifies the mapping of JPEG 2000 coded pictures into the MXF Generic Container for digital cinema applications. Building on SMPTE 422M, this standard defines the constraints and specific values required for interoperability across D-Cinema playback equipment. The JPEG 2000 profiles are defined in ISO/IEC 15444-1 Amd 1, which specifies two profiles: 2K (2048×1080) and 4K (4096×2160). These profiles were specifically designed for digital cinema, with frame-based wrapping and constrained compression parameters that ensure consistent decode performance across all compliant projectors.
The choice of JPEG 2000 as the compression technology for digital cinema was driven by several key requirements. JPEG 2000 offers lossless or visually lossless compression at the bit rates required for cinema distribution, supports the 12-bit X’Y’Z’ color space required for DCDM, and enables resolution scalability from 2K to 4K. Unlike consumer video codecs, JPEG 2000 compresses each frame independently (intra-frame coding), which is essential for frame-accurate editing and random access in cinema playback servers.
The 2K and 4K profiles use different Rsiz values in the JPEG 2000 codestream 03h for 2K and 04h for 4K. Decoders must check this value to correctly configure the decompression pipeline. Attempting to decode a 4K codestream with a 2K decoder will fail due to different tile size and component configuration parameters.
2. KLV Coding and UL Constraints
| Parameter | 2K Profile | 4K Profile | Description |
|---|---|---|---|
| Rsiz (JPEG 2000) | 03h | 04h | D-Cinema profile identifier |
| Stored Width | 2048 (max) | 4096 (max) | Horizontal pixel resolution |
| Stored Height | 1080 (max) | 2160 (max) | Vertical pixel resolution |
| Component Max Ref | 4095 | 4095 | Maximum X’Y’Z’ code value (12-bit) |
| Component Min Ref | 0 | 0 | Minimum X’Y’Z’ code value |
| Pixel Layout | D8h-0Ch-D9h-0Ch-DAh-0Ch-00h-00h | Same | X’Y’Z’ color component identification |
| Picture Essence Compression UL | 03h (byte 16) | 04h (byte 16) | JPEG 2000 codestream restriction identifier |
3. Engineering Implementation Details
The Essence Element Key for D-Cinema JPEG 2000 uses specific values in bytes 14-16 of the 16-byte UL: byte 14 = Essence Element Count (01h), byte 15 = Essence Element Type (08h for frame-wrapped JPEG 2000), byte 16 = Essence Element Number (01h). Frame-based wrapping is mandatory there shall be exactly one picture essence track per MXF file as defined in ISO 26429-3. These UL values ensure that decoders can unambiguously identify the essence type and apply the correct decoding algorithm.
The pixel layout uses D8h, D9h, and DAh for X’, Y’, Z’ components respectively. Note that these values are the ISO 7-bit character codes for X, Y, Z with the MSB set to 1. This distinguishes DCDM color components from the X, Y, Z values already defined in SMPTE 377M for other color spaces.
The RGBA Picture Essence Descriptor provides comprehensive metadata about the picture encoding. Key fields include: Sample Rate (typically {24,1} or {48,1}), Frame Layout (0 = progressive), Component Max Ref (4095 for 12-bit), Component Min Ref (0), and Gamma. The JPEG 2000 Picture Sub Descriptor carries Rsiz, image dimensions (Xsiz, Ysiz), tile size (XTsiz, YTsiz), and component sizing information including the precision bits per component. The 2K profile uses 2048×1080 image size with a single tile per frame, while the 4K profile uses 4096×2160 with multiple tiles for efficient parallel decoding.
When designing JPEG 2000 decoders for D-Cinema, the component sizing array in the sub-descriptor indicates 3 components each with 11 bits of precision (plus 1 sign bit, totaling 12 bits). This 12-bit processing path is a key differentiator from consumer JPEG 2000 decoders, which typically operate at 8-bit precision. Hardware designers must budget for the additional processing and memory bandwidth required for 12-bit pixel path processing, which effectively doubles the data bandwidth compared to 8-bit systems.
The Aspect Ratio defaults to {256,135} (approximately 1.896:1), which is the D-Cinema flat format. Other values are used when the pixel array does not fully occupy the DCDM operational level. The Video Line Map property uses an array of four Int32 values (typically 2, 4, 0, 0) to describe the first and last active lines of the video signal, which is essential for proper display synchronization.
4. Frequently Asked Questions
Q: What happens if I try to decode a 4K JPEG 2000 stream on a 2K decoder?
A: The decoder should check the Rsiz value (03h vs 04h) and reject the stream if it does not match its capability. Attempting to decode a 4K stream with a 2K decoder will produce incorrect results due to different tile sizes.
A: The decoder should check the Rsiz value (03h vs 04h) and reject the stream if it does not match its capability. Attempting to decode a 4K stream with a 2K decoder will produce incorrect results due to different tile sizes.
Q: Can I use JPEG 2000 profiles other than those defined in ISO/IEC 15444-1 Amd 1?
A: No, only the two D-Cinema-specific profiles (2K and 4K) defined in the amendment are permitted. Using other profiles would break interoperability across the D-Cinema ecosystem.
A: No, only the two D-Cinema-specific profiles (2K and 4K) defined in the amendment are permitted. Using other profiles would break interoperability across the D-Cinema ecosystem.
Q: Why is frame-based wrapping mandatory?
A: Frame-based wrapping ensures each MXF content package contains exactly one frame. This enables frame-accurate random access and simplifies editing operations compared to clip-based or multi-frame wrapping.
A: Frame-based wrapping ensures each MXF content package contains exactly one frame. This enables frame-accurate random access and simplifies editing operations compared to clip-based or multi-frame wrapping.
Q: How does the RGBA Picture Essence Descriptor identify X’Y’Z’ color space?
A: Through the PixelLayout field using the values D8h (X), D9h (Y), DAh (Z) with the MSB set to 1. This is an 8-byte array structured as 4 pairs of (component code, horizontal subsampling factor).
A: Through the PixelLayout field using the values D8h (X), D9h (Y), DAh (Z) with the MSB set to 1. This is an 8-byte array structured as 4 pairs of (component code, horizontal subsampling factor).
