System and method for estimating, synthesizing and matching noise in digital images and image sequences

1. A method for synthesizing noise in a digital image comprised of a plurality of color channels, where the noise is used to represent a synthetic texture that visually approximates a predetermined appearance, said method comprising the steps of:

(a) estimating noise statistics based on a set of spatial autocorrelation coefficients and a set of spectral correlation coefficients that correspond to the color channels, where at least one of the spatial autocorrelation coefficients is weighted by at least one of the spectral correlation coefficients, thereby providing a weighted set of autocorrelation coefficients, and where filter coefficients are obtained from the weighted set of autocorrelation coefficients; and

(b) synthesizing a synthetic texture by using the filter coefficients to filter a random noise field in each of the color channels, thereby producing output noise fields in each of the color channels that replicate the synthetic texture when combined into the digital image.

2. The method as claimed in claim 1 wherein the step (b) of synthesizing a synthetic texture comprises producing a filtered noise field in each of the channels and weighting at least one of the filtered noise fields with a portion of another filtered noise field to produce at least one of the output noise fields.

3. The method as claimed in claim 1 wherein the noise statistics estimated in step (a) are based on a set of spatial autocorrelation coefficients and a set of spectral correlation coefficients obtained from an arbitrary source.

4. The method as claimed in claim 3 wherein the arbitrary source is obtained from a plurality of predetermined media “looks” or appearances.

5. The method as claimed in claim 3 wherein the arbitrary source is originated from user interaction.

6. The method as claimed in claim 1 wherein the digital image comprises a sequence of digital images and the noise statistics estimated in step (a) are based on a set of spatial autocorrelation coefficients and a set of spectral correlation coefficients obtained from an analysis of the sequence of digital images.

7. The method as claimed in claim 6 wherein the analysis of the sequence of digital images comprises the steps of motion compensated filtering of the sequence of digital images, thereby providing filtered images, and operation on the filtered images with a noise estimation algorithm to provide the noise statistics.

8. The method as claimed in claim 7 further comprising the step of spatially masking the filtered images to identify preferred images areas for noise estimation.

9. The method as claimed in claim 1 wherein the digital image comprises a sequence of source images and another sequence of target images, and wherein the statistics estimated in step (a) are based on a set of spatial autocorrelation coefficients and a set of spectral correlation coefficients obtained from analyzing and matching the source images and the target images.

10. The method as claimed in claim 1 wherein the digital image comprises a first sequence of images and a second sequence of images, each with a different source noise level, and wherein the statistics estimated in step (a) are obtained for each sequence of images by motion filtering of each sequence, thereby providing filtered images in each sequence, and operation on the filtered images in each sequence with a noise estimation algorithm to provide the noise statistics, wherein the noise statistics in each sequence are adjusted to a similar target noise level, whereby the output noise field added to each sequence will be different but the final noise positions of the sequences will be the same.

11. The method as claimed in claim 1 wherein the synthetic texture visually approximates a desired grain appearance.