Skip to main content

Table 1 Parameters set for the original BM3D

From: Patch-based models and algorithms for image denoising: a comparative review between patch-based images denoising methods for additive noise reduction

 

Symbol

Description

Fast BM3D

Normal BM3D

    

σ≤40

σ≥40

Step 1 (hard) parameters

\(T_{\text {hard}}^{\mathrm {2D}}\)

2D transform

Bior1.5

Bior1.5

DCT

 

\(T_{\text {hard}}^{\mathrm {3D}}\)

3D transform

Haar

Haar

Haar

 

\(N_{1}^{\text {hard}}\)

Patch size

8

8

12

 

\(N_{2}^{\text {hard}}\)

3D array size

16

16

16

 

\(N_{\mathrm {step1}}^{\text {hard}}\)

Patch step

6

3

4

 

\(N_{\text {search}}^{\text {hard}}\)

Window size

25

39

39

 

\(N_{\mathrm {step2}}^{\text {hard}}\)

Window step

6

1

1

 

\(N_{\mathrm {Prev.}}^{\text {hard}}\)

Small window

3

–

–

 

β hard

Kaiser window

2.0

2.0

2.0

 

λ 2D

2D thresholding

0

0

2.0

 

λ 3D

3D thresholding

2.7

2.7

2.8

 

\(T_{\text {match}}^{\text {hard}}\)

Distance

2500

2500

5000

Step 2 (Wiener) parameters

\(T_{\text {Wiener}}^{\mathrm {2D}}\)

2D transform

DCT

DCT

DCT

 

\(T_{\text {Wiener}}^{\mathrm {3D}}\)

3D transform

Haar

Haar

Haar

 

\(N_{1}^{\text {Wiener}}\)

Patch size

8

8

11

 

\(N_{2}^{\text {Wiener}}\)

3D array size

16

32

32

 

\(N_{\mathrm {step1}}^{\text {Wiener}}\)

Patch step

5

3

6

 

\(N_{\text {search}}^{\text {Wiener}}\)

Window size

25

39

39

 

\(N_{\mathrm {step2}}^{\text {Wiener}}\)

Window step

5

1

1

 

\(N_{\mathrm {Prev.}}^{\text {Wiener}}\)

Small window

2

–

–

 

β Wiener

Kaiser window

2.0

2.0

2.0

 

\(T_{\text {match}}^{\text {Wiener}}\)

Distance

400

400

3500