Given an grayscale digital image, and a positive integer $n$, how well can we store the image at a compression ratio of $n:1$? In this paper we address the above question in extreme cases when $n>>50$ using "$\mathbf{V}$-variable image compression".