IP Core for Real-time, lossless compression in Multispectral and Hyperspectral imaging applications


Technology abstract

This IP Core technology offers lossless multispectral & hyperspectral Image Compression. When integrated on-board a satellite, aircraft or a UAV, it can be the key enabling technology for next-generation precision agriculture, surveillance, security, environmental monitoring and mineralogy remote sensing applications. The tech provider, a leading research laboratory at the University of Athens, is looking for partners that will integrate this technology in their systems. 

Technology Description

Sensors used in Multispectral and Hyperspectral imaging capture the spectral fingerprint of an object, thus providing a unique spectral signature enabling accurate object detection and classification. The high-resolution and the high-speed needed in such applications create a huge volume of data that the imager payloads cannot handle due to the limited on-board/aerial storage resources and the limited speed of downlink bandwidth, making image compression one of the most challenging remote-sensing data processing tasks.
The CCSDS 123.0-B-1 IP Core offered achieves state-of-the-art data-rate performance enabling real-time lossless compression of Multispectral & Hyperspectral images for on-board/aerial usage.
The technology provider has significant expertise in high-performance on-board data handling and especially high performance image/data compression either firmware (FPGA) or software in accordance to the European Cooperation for Space Standardisation (ECSS) Space product assurance: ASIC and FPGA development (ECSS-Q-ST-60-02C).

Innovations & Advantages

Existing field-programmable gate array (FPGA) implementations of lossless Multispectral & Hyperspectral image compression based on CCSDS 123.0-B-1 (or NASA Fast Lossless (FL) on which the recommended standard is based), using either Band-Interleaved-by-Pixel (BIP) or Band-SeQuential (BSQ) orderings, achieve up to 40 MSamples/sec (NASA JPL) when targeting Xilinx Virtex-5 FPGA as reference technology, which may not be enough for today’s and future high data-rate Multispectral and Hyperspectral imager payloads in order to provide real-time compression.
The high-performance architecture of the CCSDS 123.0-B-1 IP Core offered, leverages BIP encoding order which minimizes data dependencies and enables pipelining that yields a high data-rate performance implementation which makes real-time Multispectral and Hyperspectral compression feasible.
When implemented in a Xilinx Virtex-5 FPGA, it achieves state-of-the-art data-rate performance with a maximum throughput of 120 MSamples/s (1.9 Gbps @ 16-bits) using about 2000 slices and 22 BRAMs.

Further Information

The cutting-edge technology CCSDS 123.0-B-1 Lossless Multispectral & Hyperspectral Image Compression IP core has the following features:
i) pixel dynamic range up to 16 bits
ii) image dimensions up to 4096 pixels for each dimension (Nx, Ny, Nz)
iii) BIP encoding order which minimizes data dependencies and enables pipelining that yields a high data-rate performance implementation
iv) number of prediction bands P=3 since a higher number of prediction bands does not yield any improvement in compression effectiveness
v) full and reduced prediction modes
vi) neighbour and column oriented local sum calculation
vii)  sample-adaptive encoding (block-adaptive encoding as an option leveraging an existing CCSDS 121.0-B-2 IP Core),
viii) simple, parallel FIFO-based streaming I/O interface
ix) memory mapped configuration registers accessible by standard interfaces (e.g. AMBA slave APB).
x) At the system level, a memory controller is required to provide access to external DRAM (e.g. DDR2/DDR3) to buffer the neighbourhood samples. The external DRAM memory storage requirements are modest; its size should be enough to hold one spatial-spectral slice of data (Nx x Nz).
The performance of the CCSDS 123.0-B-1 IP core has been extensively verified using RTL simulation and FPGA-in-the-loop based verification using the corpus of Hyperspectral and Multispectral test images and test patterns. The ESA software implementation in C was used as a golden reference model. Furthermore, the IP Core was validated using a Xilinx development board as a hardware demonstrator.

Current and Potential Domains of Application

The CCSDS 123.0-B-1 IP core can be the key-enabling technology for real-time Lossless Multispectral & Hyperspectral Image Compression for next-generation remote-sensing applications such as: precision agriculture, surveillance, security, environmental monitoring and mineralogy.
The advantages that this IP core offers is essential is such applications where the huge volume of remote sensing data from high-resolution and high-speed Multispectral and Hyperspectral imager payloads along with the limited on-board storage resources and downlink bandwidth pose a great challenge.
Efficient on-board lossless Multispectral and Hyperspectral data compression is a feasible solution to meet satellite’s limited downlink capabilities while it also improves signature extraction, object recognition and feature classification capabilities by providing exact reconstructed data on constrained downlink resources.
Real-time Multispectral & Hyperspectral image compression is required in order to support immediate decision making in critical circumstances.