Digital image processing method for edge shaping

• US 6,611,627 B1
• Filed: 04/24/2000
• Issued: 08/26/2003
• Est. Priority Date: 04/24/2000
• Status: Expired due to Term

• Current Assignee: Monument Peak Ventures, LLC (Dominion Harbor Enterprises, LLC)
• Original Assignee: Eastman Kodak Company
• Inventors: LaRossa, Gina N., Lee, Hsien-Che
• Primary Examiner(s): Do, Anh Hong
• Application Number:
  US09/556,666
• Time in Patent Office: 1,219 Days
• Field of Search: 382/240, 382/274, 382/254, 382/199, 382/224, 382/131, 382/132, 382/248, 382/276, 706/20, 348/125
• US Class Current: 382/240
• CPC Class Codes: G06T 2207/20064 Wavelet transform [DWT]
• CPC Class Codes: G06T 2207/20192 Edge enhancement; Edge preservation
• CPC Class Codes: G06T 5/10 using non-spatial domain filtering
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening

Abstract

A digital image processing method, includes the steps of: transforming the digital image using an edge sensitive wavelet transform to produce a plurality of wavelet coefficients at various resolutions and a residual image; modifying the wavelet coefficients as a function of the image gradient and its rate of change at a resolution corresponding to the respective wavelet coefficient; and inverse transforming the modified wavelet coefficients and the residual image to produce a processed digital image.

Claims

1. A digital image processing method, comprising the steps of:

a) transforming a digital image using an edge sensitive wavelet transform to produce a plurality of wavelet coefficients at various resolutions and a residual image;

b) modifying the wavelet coefficients as a function of the rate of change of the image gradient at a resolution corresponding to the respective wavelet coefficient, wherein the rate of change comprises an image gradient curvature that is computed by applying a Laplacian operator to the image gradient, thereby generating a modification factor that is invariant to scaling of the image; and

c) inverse transforming the modified wavelet coefficients and the residual image to produce a processed digital image.

2. The method claimed in claim 1, wherein the edge sensitive wavelet transform is a Mallat-Zhong edge-wavelet.

3. The method claimed in claim 1, wherein the rate of change of the gradient is determined as the normalized Laplacian of the gradient of the image at the respective resolution.

4. The method claimed in claim 1, wherein the function of the rate of change of the gradient at resolution i, has the functional form:

5. The method claimed in claim 4, wherein:

0.8<

b_i<

1.5;

1.0<

A_i<

3.0;

the metric is log exposure where 2.0<

β

_i<

4.0.

6. The method claimed in claim 1, wherein the function of the gradient at resolution i has the functional form:

$f_a=\frac{G_i^n}{G_i^n+{\left({\mathrm{\alpha <br><br>}}_i<br><br>{\mathrm{\gamma <br><br>}}_i\right)}^n},$

where α

_i and n are chosen based on the type of noise present in the image at resolution i, γ

_i represents the gradient having the highest count in a gradient histogram, and G_i={square root over (H_i²+V_i²)}, $H_i=\frac{\partial <br><br>B_{i-1}}{\partial <br><br>x},\text{\hspace{1em}<br><br>}<br><br><br><br>V_i=\frac{\partial <br><br>B_{i-1}}{\partial <br><br>y},$

and B_i is the digital image at resolution i.

7. The method claimed in claim 1, wherein the wavelet coefficients are modified by applying a wavelet coefficient-scaling mask m_i(x,y).

8. The method claimed in claim 7, wherein the wavelet coefficient-scaling mask is generated by:

a) forming an edge map by locating directional local maximum points of the image gradient and labeling pixels (x_e,y_e) in the edge map as edge pixel if the gradient is a local maximum along the gradient direction;

b) computing the values of the wavelet coefficient-scaling mask m_i(x_e,y_e) at edge pixels (x_e,y_e) as the product of an attenuation factor f_a and a boost factor f_b; and

c) computing the values of the wavelet coefficient scaling mask at non-edge pixels as the average of contributions from neighboring edge pixels.

9. The method claimed in claim 8, wherein the attenuation factor f_a is implemented with the following functional form $f_a=\frac{G_i^n}{G_i^n+{({\alpha }_{}}^{}}$

i

γ

i)n,

where γ

_i represents the gradient having the highest count in a gradient histogram, (α

_i and n are chosen based on the type of noise present in the image at that resolution, and G_i={square root over (H_i²+V_i²)}; and

the boost factor f_b is implemented with the following functional form

10. A computer program product for a digital image processing method comprising:

a computer readable storage medium having a computer program stored thereon for performing the steps of;

a) transforming a digital image using an edge sensitive wavelet transform to produce a plurality of wavelet coefficients at various resolutions and a residual image;

b) modifying the wavelet coefficients as a function of the rate of change of the image gradient at a resolution corresponding to the respective wavelet coefficient, wherein the rate of change comprises an image gradient curvature that is computed by applying a Laplacian operator to the image gradient, thereby generating a modification factor that is invariant to scaling of the image; and

c) inverse transforming the modified wavelet coefficients and the residual image to produce a processed digital image.