A New Algorithm for the On-Board Compression of Hyperspectral Images

Abstract

Hyperspectral sensors are able to provide information that is useful for many different applications. However, the huge amounts of data collected by these sensors are not exempt of drawbacks, especially in remote sensing environments where the hyperspectral images are collected on-board satellites and need to be transferred to the earth's surface. In this situation, an efficient compression of the hyperspectral images is mandatory in order to save bandwidth and storage space. Lossless compression algorithms have been traditionally preferred, in order to preserve all the information present in the hyperspectral cube for scientific purposes, despite their limited compression ratio. Nevertheless, the increment in the data-rate of the new-generation sensors is making more critical the necessity of obtaining higher compression ratios, making it necessary to use lossy compression techniques. A new transform-based lossy compression algorithm, namely Lossy Compression Algorithm for Hyperspectral Image Systems (HyperLCA), is proposed in this manuscript. This compressor has been developed for achieving high compression ratios with a good compression performance at a reasonable computational burden. An extensive amount of experiments have been performed in order to evaluate the goodness of the proposed HyperLCA compressor using different calibrated and uncalibrated hyperspectral images from the AVIRIS and Hyperion sensors. The results provided by the proposed HyperLCA compressor have been evaluated and compared against those produced by the most relevant state-of-the-art compression solutions. The theoretical and experimental evidence indicates that the proposed algorithm represents an excellent option for lossy compressing hyperspectral images, especially for applications where the available computational resources are limited, such as on-board scenarios.

Towards the Concurrent Execution of Multiple Hyperspectral Imaging Applications by Means of Computationally Simple Operations

“…On this basis, the extraction of the most representative pixels of materials present in a scene permits performing dimensionality reduction and thus, to compress the HSIs as analyzed in [41] for the lossy compression of HSIs. Considering that a HSI is composed of pe pixels with nb spectral bands, if these pe pixels are projected onto a subset of the p most different pixels within the scene, the original pe pixels can be represented as a linear combination of their projections onto those p pixels.…”

“…After the extraction of the p c characteristic pixels, e c n , and projection vectors, v c n , using the set of core operations, our proposal follows the same methodology as [41] to slightly increase the compression ratio at a very low computational cost and without introducing further losses of information. To do this, outputs E c , V c and µ are independently processed in two stages, named Preprocessing and Entropy Coding in Figure 1.…”

“…In this stage, each single output vector µ, e c n and v c n is independently coded using a Golomb-Rice coder [51]. For more information, we encourage the reader to see [41].…”

“…Moreover, in order to evaluate our proposal in real scenarios, we have employed a new data set using real hyperspectral data collected by one of our unmanned aerial platforms. In any case, we have tested the goodness of the Gram-Schmidt method using conventional hyperspectral data sets in previous works for lossy compression and anomaly detection in [41,43], respectively. Despite the set of core operations proposed in this work has been optimized for allowing the concurrent execution of both processes, the obtained results are exactly the same that if both were independently executed using the already published algorithms.…”

“…Concretely, the proposed method uses a modified version of the Gram-Schmidt orthogonalization process [39,40] that allows extraction of useful information for many different hyperspectral imaging applications while at the same time, ensuring a low computational complexity and a high level of parallelism. The advantages of this set of core operations have already been tested using different algorithms specifically developed for unmixing applications [40], hyperspectral lossy compression [41] and anomaly detection [42,43]. Additionally, some of these algorithms have been successfully implemented into parallel computing devices, such as LPGPUs mounted on a UAV [20] and next-generation space-grade FPGAs [44], achieving real-time performance in both cases.…”