The Multi-scale CLEAN algorithm has also been designed to improve the performance of CLEAN for extended sources. It is a straightforward extension of CLEAN that models the sky brightness by the summation of blobs of emission having different size scales. It is equivalent to simultaneously deconvolve images obtained with different synthesized beams derived from the highest resolution one by convolution kernels. This algorithms works simultaneously in a range of specified scales. Multi-scale CLEAN can produce good images with a loop gain of 0.5 or even higher.
The implementation of Multi-scale CLEAN in IMAGER slightly differs from that of CASA . It is less optimized in terms of speed, but uses a better convergence scheme in which the scale chosen at each iteration is the one with best signal to noise ratio. Accordingly, it is more stable. Only 3 scales are used so far in IMAGER, with a size ratio controlled by CLEAN_SMOOTH.