Truncated Nuclear Norm Minimization for Image Restoration Based on Iterative Support Detection

Figure 5

We present the process of SVE in LRISD-ADMM about three images in Figure 6. At the same time, we note that the first and the second singular values are much larger than others, as well as the values of . To make the results more clear, we omit the first and the second singular values and in each figure. We can find the observed estimated are 7, 9, 8. Compared to the best , which are 8, 10, 7, estimated is approximately equivalent to the best .

(a) Door: the first outer iteration of LRISD

(b) Door: the second outer iteration of LRISD

(c) Door: the third outer iteration of LRISD

(d) Window: the first outer iteration of LRISD

(e) Window: the second outer iteration of LRISD

(f) Window: the third outer iteration of LRISD

(g) Sea: the first outer iteration of LRISD

(h) Sea: the second outer iteration of LRISD

(i) Sea: the third outer iteration of LRISD