A computer program that preprocesses multispectral image data has been developed to provide the Mars Exploration Rover (MER) mission with a means of exploiting the additional correlation present in such data without appreciably increasing the complexity of compressing the data. When used in conjunction with ICER, a previously developed image-data-compression program, this program enables improved compression of multispectral images, compared to that achievable by use of ICER alone. As such, it is a straightforward means of achieving much of the gain possible from exploiting spectral correlation. This preprocessor software accommodates up to seven images that are different spectral bands of the same scene. The software performs an approximate discrete cosine transform (DCT) pixelwise across the spectral bands. The software is written for speed; in particular the DCT operation performs only integer operations (producing integer output) and uses multiplications sparingly. Separate code is used for each possible number of spectral bands, including numbers for which fast DCT functions are not normally implemented. The DCT output is scaled so that, if the original images have a bit depth of at most 12, the transformed images are guaranteed to have a dynamic range appropriate for compression by the ICER software on the MER rovers. The resulting transformed bands are compressed individually by ICER. To reconstruct the images, the transformed images are first decompressed by use of the decompressor for ICER, then the resulting reconstructed images are passed to an inverse-DCT subprogram, which reconstructs the various spectral bands.
This work was done by Matthew Klimesh and Aaron Kiely of Caltech for NASA's Jet Propulsion Laboratory. For further information, access the Technical Support Package (TSP) free on-line at www.techbriefs.com/tsp under the Software category.
This software is available for commercial licensing. Please contact Karina Edmonds of the California Institute of Technology at (626) 395- 2322. Refer to NPO-40835.
This Brief includes a Technical Support Package (TSP).
Pre-Processor for Compression of Multispectral Image Data
(reference NPO-40835) is currently available for download from the TSP library.
Don't have an account?
Overview
The document outlines the development and application of a software pre-processor designed to enhance the compression of multispectral images, particularly for the Mars Exploration Rover (MER) mission. This pre-processor works in conjunction with the previously developed ICER (Image Compression for Efficient Representation) image compressor, enabling improved compression performance compared to using ICER alone.
The pre-processor leverages the additional correlation present in multispectral images, which consist of multiple spectral bands captured from the same scene. It performs an approximate discrete cosine transform (DCT) across these bands, allowing for more efficient compression. The transformed bands are then compressed individually using ICER, and the reconstruction process involves decompressing these bands and applying an inverse DCT to produce the final images.
The document highlights several key features of the pre-processor:
-
Compatibility: It can handle up to seven spectral bands, accommodating the needs of the MER project. Modifications for larger numbers of bands are straightforward.
-
Efficiency: The DCT function is optimized for speed, utilizing integer operations and minimizing the number of multiplications, which is crucial for real-time applications.
-
Compression Quality: The pre-processor is designed to achieve a balance between compression efficiency and image quality. However, it is noted that the DCT used is lossy, meaning that exact reconstruction of the original images is not possible.
-
Performance Metrics: The document presents performance data through rate-distortion curves, comparing the effectiveness of applying ICER directly to each band versus using the DCT pre-processor. The results indicate that the DCT significantly improves performance, often approaching the optimal results achieved with the more complex Karhunen-Loève Transform (KLT).
-
Signal Energy Considerations: It is noted that certain bands, such as L2, may have higher signal energy, which can affect compression performance. The document emphasizes that best results are obtained when the spectral bands have similar energy levels.
Overall, the pre-processor represents a practical solution for enhancing multispectral image compression without significantly increasing computational complexity or development time. While it is not the ultimate solution for all compression needs, it provides a valuable tool for the MER mission and potentially other applications in aerospace and beyond.
