Compression algorithm for Multi Element Telescope for Imaging and Spectroscopy (METIS)

The compression algorithm defined for METIS (Multi Element Telescope for Imaging and Spectroscopy) arises from the standard CCSDS 123.0-r-1, that has been modified and adapted to the mission purposes, and integrated with other pieces of software to let the compressor work in the most efficient way with the expected acquisitions of the sensor. The major modification is the insertion in the prediction loop of a uniform scalar quantizer, extending the standard to a near-lossless version; in addition a local decoder has been added as well, in order to keep a local copy of the dequantized residuals to allow a correct reconstruction at the decoder side. A lossy compression can even be executed in a variable-quality way, meaning that it is possible to change the quantization step size among successive image lines. The ability of the original software to process three-dimensional images has been kept but adapted to the mission needs: instead of considering wavelength, consecutive acquisitions are collected together to build up the 3D cube, so that time becomes the third dimension; and since solar acquisitions change really slowly in time, the effectiveness of this adjustment works very well and prediction of the current pixels becomes much more accurate if considering the previous acquisitions ones. Further, a pre-processing routine has been developed to exploit the geometry of the images; it consists in a re-mapping of the pixels in order to take advantage of the radial structure of solar acquisitions, through a function that has been named “radialization”. It receives the standard image as input, and computes for every pixel the distance and the angle with respect to the center; these become the two new coordinates, as it happens when switching from a Cartesian system to a polar one. The triangular-shaped output is then centered and padded in order to keep a rectangular structure, and matrices for the two dimensions are kept, so that the whole piece of code can be executed only once, and the “radialized” image can be then obtained by a simple mapping using these structures, resulting in a really light operation from a computational point of view; a further advantage can be identified in the lack of interpolation among pixels, so that eventually, the compression of the image, or better of a section of it, can occur losslessly. Radialization also simplifies a possible selection of areas of interest of the image: for example it would be possible to keep the nearest solar corona area coded losslessly, and decreasing linearly the quality of the reconstruction in a radial sense by successive circular corona-shaped structures, by using variable lossy compression for consecutive radialized image lines.