Noise reduction in color digital images using pyramid decomposition

1. A method of removing noise from a color digital image, comprising the steps of:

(a) receiving an original color digital image including a plurality of pixels represented in a primary-color space;

(b) producing at least one residual digital image and at least one base digital image from the original color digital image, the base digital image having a lower spatial resolution then the original color digital image;

(c) producing a noise reduced digital image by removing noise from the residual image and the base digital image with a noise reduction filter and combining the noise reduced base digital image with the noise reduced residual image to produce a reconstructed digital image having reduced noise;

(d) transforming the reconstructed digital image into a luminance-chrorninance color space;

(e) repeating steps (b) and (c) on the luminance-chrominance digital image to produce a reconstructed luminance-chrominance digital image; and

(f) transforming the reconstructed luminance-chrominance digital image into a noise reduced digital image in the original primary color space.

2. The method of claim 1 wherein the primary-color space is red, green, and blue (RGB).

3. The method of claim 2 wherein the luminance-color space is defined by the transform

{

⁢

Y

=

G

⁢

C

1

=

2

⁢

G

-

R

-

B

4

⁢

C

2

=

B

-

R

2

.

wherein

Y is luminance;

C1 is first chrominance channel; and

C2 is second chrominance channel.

4. The method of claim 1 wherein the residual and base digital images are the components of a Laplacian pyramid image decomposition.

5. The method of claim 1 wherein element (d) includes a logarithmic transform

y= 2500[log 10 ( x+ 100)−2]

wherein

y is the transformed code value; and

x is the original code value.

6. The method of claim 1 wherein the noise reduction employs a direction blurring operation.

7. The method of claim 1 wherein the noise reduction filter includes a sigma filter.