Image processing for synthesis imaging of mingantu spectral radioheliograph (MUSER)

Abstract

MingantU SpEctral Radioheliograph (MUSER) can generate the images of the Sun with high time, frequency and spatial resolutions. It employs the aperture synthesis (AS) principle to image the Sun with plentiful solar radio activities. Different from general imaging system, AS records sparse Fourier components of the spatial image of the Sun by recurring to electromagnetic interference imaging principle. However, due to the limited number of antennas, the recorded Fourier components is extremely sparse, which results in very blurring images. This problem is equivalent to convoluting an image with a Gaussian smoothing filter in spatial domain. Accordingly, one can recover an image from its burred version by inverse operation of convolution, namely deconvolution, which was widely known as CLEAN algorithm. This algorithm however does not perform well on solar images characterized by extended source instead point source. In this paper, a new method based on compressed sensing (CS) is proposed to replace CLEAN for imaging or preceded by CLEAN. It describes itself a standard optimization function constrained by sparseness. Specifically, it adopts structural group dictionary to represent solar images for exploring both local sparsity and nonlocal self-similarity. We also investigate image reconstruction of MUSER in a wider range by employing several state-of-the-art image deburring methods. The performance analysis reveals that the proposed method contributes image quality improvement of MUSER markedly beyond the other methods.