Compression of hyperspectral images containing a sub-pixel target

Hyperspectral (HS) image sensors measure the reflectance of each pixel at a large number of narrow spectral bands, creating a three-dimensional representation of the captured scene. Since the HS image (HSI) consumes a great amount of storage space and transmission time, it would be desirable to reduce the image representation to the extent possible using a compression method, which corresponds to the usage and processing of the image. Many compression methods have been proposed, aiming at different applications and fields. This research focuses on lossy compression of images that contain sub-pixel targets. This target type requires minimum compression loss over the spatial dimension, in order to preserve the target, and the maximum possible spectral compression that would still enable target detection. For this target type, we propose the PCA-DCT (principle component analysis followed by discrete cosine transform) compression method. It combines the PCA ability to extract the background from a small number of components, with the individual spectral compression of each pixel of the residual image, using quantized DCT coefficients. The compression method is kept simple for fast processing and implementation, and considers lossy compression only on the spectral axis. It achieves compression ratio of over 20, while using only spectral compression (before applying spatial compression and bit-stream-encoding). The popular RX (Reed Xiaoli) algorithm and the improved quazi-local RX (RX_QLC) are used as target detection methods. The detection performance is evaluated using ROC (receiver operating characteristics) curve generation. The proposed compression method shows improved detection performance, compared to the detection performance of the original image, and of two other compression methods: PCA-ICA and band decimation.