A New Boosting Algorithm for Shrinkage Curve Learning

Figure 5

Visual and PSNR comparisons of several algorithms of cropped region from a noisy image. The first column is a noisefree image, the second column is a noisy image, whose corresponding noise level is below the image, the third and fourth columns are the results of curve denoising trained by using accurate and estimated noise levels, respectively, and the fifth column shows the denoising results of the proposed algorithm. (a) Lena. (b) σ = 5. (c) σ₀  = 36.82. (d)  = 36:96. (e) Ours = 37.91. (f) House. (g) σ = 10. (h) σ₀ = 32.88. (i)  = 32.93. (j) Ours  = 34.71. (k) Peppers. (l) σ = 20. (m) σ₀ = 29.00. (n) = 29.01. (o) Ours = 29.60. (p) Fingerprint. (q) σ = 25. (r) σ₀ = 26.77. (s)  = 26.79. (t) Ours = 27.35. (u) Montage. (v) σ = 50. (w) σ₀ = 24.39. (x) = 24.30. (y) Ours = 25.49.

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

(i)

(j)

(k)

(l)

(m)

(n)

(o)

(p)

(q)

(r)

(s)

(t)

(u)

(v)

(w)

(x)

(y)