Claims
- 1. A method for noise estimation and reduction in digital images comprising the following steps:
- providing a digital image having a range of code values and having signal dependent noise;
- measuring signal dependent noise in said digital image (FIG. 1 Block 10), with a Noise Estimation System (20) which calculates a noise standard deviation estimate corresponding to each code value found in said digital image;
- using said noise estimates (30) to configure a Noise Removal System (40);
- processing said digital image (10) with the above configured Noise Removal System (40) in order to generate a Noise Reduced Version of said digital image (50).
- 2. A method to Estimate Signal Dependent Noise in Images (20) comprising the following steps:
- providing a digital image having signal dependent noise;
- processing said digital image (10,60) with a Gradient Estimation System (FIG. 2, Block 70), in order to generate a Gradient Image (130), and storing said Gradient Image in Memory (90);
- processing said Gradient Image (60) with a Threshold Segmentation System (80), in order to generate a plurality of Mask Images (100, 110, . . . , 120), each representing a specific range of pixel code values in said digital image (60), and storing said Mask Images in memory (90); processing said Gradient Image (130) with each of the plurality of Mask Images (100, 110, . . . , 120), in order to generate a plurality of Histograms (150, 160, . . . , 170), and storing said histograms in memory (90);
- processing the plurality of said Histograms (150, 160, . . . , 170) through a Gradient Histogram Threshold Calculation System (180) in order to generate a plurality of scalar Thresholds (200, 210, . . . , 220) and storing said scalar Thresholds in memory (90); and
- processing said digital image (60), the plurality of Mask Images, and the plurality of scalar Thresholds with a Noise Measurement System (190) in order to generate a plurality of Signal Dependent Noise Estimates (230, 240, . . . , 250), each Signal Dependent Noise Estimate corresponding to a specific band of code values in said digital image (60), and storing said Signal Dependent Noise Estimates in memory (90).
- 3. The method of claim 2 wherein processing with a Noise Measurement System step includes the following steps:
- the Gradient Image, the plurality of Mask Images, and the plurality of Scalar Thresholds are processed together by a gradient segmentation system to produce a new set of Mask Images and Masks;
- the digital image is processed by an image estimation system to produce an estimated image;
- the digital image and the estimated image are processed together by an image difference system to produce a difference image including relevant pixels; and
- for each member of the new set of Mask Images, the relevant pixels in the difference image are operated on by a statistics system to produce a plurality of noise estimates.
- 4. The method of claim 2 wherein said Gradient Histogram Threshold Calculation System includes the steps of:
- each of said plurality of histograms is processed for robustness by a histogram smoothing system to produce a plurality of s smoothed histograms;
- each of said histograms is processed by a peak detection system to produce a plurality of scalar peak locations;
- each scalar peak location is processed by a peak adjustment system to produce said plurality of scalar thresholds.
CROSS REFERENCE TO RELATED APPLICATION
Reference is made to and priority claimed from U.S. Provisional Application Ser. No. U.S. 60/014,863, filed Apr. 4, 1996, entitled APPARATUS AND METHOD FOR SIGNAL DEPENDENT NOISE ESTIMATION AND REDUCTION IN DIGITAL IMAGES.
US Referenced Citations (5)
Number |
Name |
Date |
Kind |
5091972 |
Kwon et al. |
Feb 1992 |
|
5533149 |
Kaplan et al. |
Jul 1996 |
|
5563963 |
Kaplan et al. |
Oct 1996 |
|
5724454 |
Shimazu et al. |
Mar 1998 |
|
5764307 |
Ozcelik et al. |
Jun 1998 |
|
Foreign Referenced Citations (1)
Number |
Date |
Country |
0 524 969 |
Dec 1993 |
EPX |