Multispectral analysis of solar EUV images: linking temperature to morphology

Context: Solar images taken simultaneously at different wavelengths in the EUV are widely used for understanding structures such as flares, coronal holes, loops, etc. The line-of-sight integration and the finite spectral resolution of EUV telescopes, however, hinders interpretation of these individual images in terms of temperature bands. Traditional approaches involve simple visualisation or explicit modelling. We take a more empirical approach, using statistical methods. Aims: The morphology of solar structures changes with the wavelength of observation and, therefore, with temperature. We explore the possibility of separating the different solar structures from a linear combination of images. Methods: Using a blind source separation approach, we build a new set of statistically independent ''source'' images from the original EUV images. Two techniques are compared: the singular value decomposition and independent component analysis. Results: The source images show more contrast than the original ones, thereby easing the characterisation of morphological structures. A comparison with the differential emission measure shows that each source image also isolates structures with specific emission temperatures.