Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
ISO 26868:2009 Space Data and Information Transfer Systems — Image Data Compression
CCSDS-Adapted Lossless-to-Lossy Image Compression Algorithm for Spaceborne Payload Instruments
1. Introduction and Scope
ISO 26868:2009 adopts the CCSDS 122.0-B-1 “Image Data Compression” Recommended Standard as an International Standard, establishing a compression algorithm for two-dimensional digital spatial image data from payload instruments on space missions. The standard addresses one of the most critical challenges in space data systems: reducing the volume of digital image data to conserve transmission bandwidth, buffering and storage capacity, and data-transmission time.
Space missions generate enormous volumes of image data from scientific instruments — a single high-resolution multispectral imager can produce gigabits of data per second. Without efficient compression, the downlink bandwidth becomes the bottleneck, limiting scientific return from valuable space assets.
The standard, developed by the Consultative Committee for Space Data Systems (CCSDS) and adopted by ISO/TC 20/SC 13, supports both lossless and lossy compression modes, giving mission designers the flexibility to trade image quality against data rate depending on the scientific requirements and available link capacity.
2. Compression Algorithm Architecture
2.1 Two-Dimensional Discrete Wavelet Transform (DWT)
The compression algorithm employs a two-dimensional discrete wavelet transform as its first processing stage. The DWT decomposes the image into subbands at multiple resolution levels, separating low-frequency (approximation) information from high-frequency (detail) components in three orientations: horizontal, vertical, and diagonal. This decomposition concentrates image energy into a small number of coefficients, enabling efficient compression through subsequent bit-plane encoding.
| Component | Function | Output Characteristic |
|---|---|---|
| 2D DWT (9/7 or 9/3 Float/Integer) | Multi-resolution decomposition | 4 subbands per level (LL, HL, LH, HH) |
| DC Coefficient Quantization | Lossy compression control | User-selectable bit depth reduction |
| Bit-Plane Encoder (BPE) | Entropy coding of wavelet coefficients | Embedded bitstream, progressive by quality |
| Segment Assembly | Packetization for space link | Fixed-size segments with headers |
2.2 Bit-Plane Encoding (BPE)
The BPE operates on 3×3 coefficient blocks grouped recursively across subbands, encoding bit-planes from most significant bit (MSB) to least significant bit (LSB). This structure produces an embedded bitstream that can be truncated at any point, enabling progressive decoding and precise rate control — a crucial feature for space links where available bandwidth may vary during a pass.
The embedded bitstream architecture means a single compressed file can serve multiple purposes: a quick-look version from the first few bits, a high-quality version for detailed analysis, and everything in between — all without re-encoding.
3. Space-Specific Design Features
3.1 Error Resilience and Segment Structure
Unlike terrestrial image compression standards (JPEG, JPEG 2000), ISO 26868 is designed for the high-radiation, high-error-rate environment of space communications. The compressed data is organized into independent segments of specified byte length, each with its own header containing segment type, length, and optional checksum information. This segmentation ensures that data corruption in one segment does not affect the decodability of others — a critical requirement for deep-space missions where retransmission may be impractical.
| Segment Type | Content | Error Recovery |
|---|---|---|
| Header-only | Image dimensions, DWT levels, compression parameters | Full — contains no coded data |
| Header + data | Partial header repeat + encoded coefficients | Partial — decoder continues at segment boundary |
| Data-only | Continuation of encoded coefficients | Partial — sync losses limited to one segment |
3.2 Compression Ratio Capabilities
The standard supports a wide range of compression ratios. In lossless mode, typical space instrument images achieve compression ratios of 2:1 to 3:1. In lossy mode, compression ratios of 5:1 to 20:1 are practical with good visual quality, while even higher ratios (50:1 or more) are possible for quick-look applications. The DWT-based approach avoids the blocking artefacts characteristic of DCT-based compression (JPEG), providing superior image quality at equivalent bit rates.
4. Engineering Design Insights
For aerospace engineers implementing ISO 26868 in onboard processing systems, several design considerations are paramount. The standard achieves compression performance comparable to JPEG 2000 with significantly lower complexity, making it attractive for FPGA implementation in space-qualified hardware:
Computational complexity: The 9/7 wavelet filter (lossy mode) requires floating-point arithmetic and more memory, while the 9/3 integer filter (lossless mode) uses only integer operations. FPGA implementations of the integer DWT can achieve real-time processing for high-resolution imagers with modest power budgets. Many space-qualified FPGAs now include dedicated DSP slices optimized for DWT operations, significantly reducing the gate count penalty compared to software implementations on radiation-hardened processors.
Rate control strategy: The three-segment-per-block encoding approach allows precise bit allocation. Engineers can implement feedback loops that adjust the quantization step size based on the achieved compression ratio of previously encoded segments, ensuring the output fits within the telemetry allocation. For missions with highly variable downlink capacity — such as deep-space probes or Martian orbiters — the progressive bitstream nature allows graceful degradation of image quality without requiring re-encoding.
Memory optimization: The standard requires storing wavelet coefficients for multi-level decomposition before encoding. For resource-constrained spacecraft, a tile-based approach (processing the image in strips) can dramatically reduce onboard memory requirements while maintaining compression efficiency. This is particularly relevant for small satellite platforms where memory resources are severely constrained and every kilobyte must be carefully allocated.
Engineers must carefully verify that the implementation handles the full range of possible input data values without overflow or underflow — space-grade FPGAs and processors often lack the headroom of terrestrial computing hardware, and a single overflow can corrupt an entire image segment.
5. FAQs
Q1: How does ISO 26868 differ from JPEG 2000?
Both use DWT and bit-plane encoding, but ISO 26868 is optimized for space applications with fixed-size segments, simplified entropy coding, reduced complexity for radiation-hardened implementation, and specific provisions for the space link environment.
Both use DWT and bit-plane encoding, but ISO 26868 is optimized for space applications with fixed-size segments, simplified entropy coding, reduced complexity for radiation-hardened implementation, and specific provisions for the space link environment.
Q2: Can the standard be used for non-space applications?
Yes — the algorithm is effective for any two-dimensional image data, particularly scientific and medical imaging where the progressive quality and lossless-to-lossy flexibility are valuable.
Yes — the algorithm is effective for any two-dimensional image data, particularly scientific and medical imaging where the progressive quality and lossless-to-lossy flexibility are valuable.
Q3: What are the memory requirements for implementation?
For a 1024×1024 image compressed to segment sizes matching the CCSDS space packet protocol, approximately 8–12 MB of working memory is typical, though strip-based processing can reduce this substantially.
For a 1024×1024 image compressed to segment sizes matching the CCSDS space packet protocol, approximately 8–12 MB of working memory is typical, though strip-based processing can reduce this substantially.
Q4: Does the standard support hyperspectral image data?
The standard covers two-dimensional spatial image data only. For hyperspectral data (3D cube), each spectral band is typically compressed independently, though work on inter-band prediction is ongoing in CCSDS.
The standard covers two-dimensional spatial image data only. For hyperspectral data (3D cube), each spectral band is typically compressed independently, though work on inter-band prediction is ongoing in CCSDS.
