3D WAVELET-BASED COMPRESSION OF HYPERSPECTRAL IMAGERY
James E. Fowler, Justin T. Rucker · 2007
Since hyperspectral imagery is generated by collecting hundreds of contiguous bands, uncompressed hyperspectral imagery can be very large, with a single image potentially occupying hundreds of megabytes. For instance, the Airborne Visible InfraRed Imaging Spectrometer (AVIRIS) sensor is capable of collecting several gigabytes of data per day. Compression is thus necessary to facilitate both the storage and the transmission of hyperspectral images. Since hyperspectral imagery is typically collected on remote acquisition platforms, such as satellites, the transmission of such data to central, often terrestrial, reception sites can be a critical issue. Thus, compression schemes oriented to the task of remote transmission are becoming increasingly of interest in hyperspectral applications. Although there have been a number of approaches to the compression of hyperspectral imagery proposed in recent years—prominent techniques would include vector quantization (VQ) (e.g., [1, 2]) or principal component analysis (PCA) (e.g., [3, 4]) applied to spectral pixel vectors, as well as 3D extensions of common image-compression methods such as the discrete cosine transform (DCT) (e.g., [5])—most of the approaches as proposed are not particularly well-suited to the image-transmission task. That is, in many applications involving the communication of images, progressive transmission is desired in that successive reconstructions of the image are possible. In such a scenario, the receiver can produce a low-quality representation of the image after having received only a small portion of the transmitted bitstream, and this “preview” representation can be successively refined in