Estimating the clutter of digital images

1. A method for determining an estimated clutter level of an input digital image, the method being performed at least in part using a data processor, comprising:

designating a set of reference features;

partitioning the input digital image into small sub-images;

determining a set of image features by automatically analyzing the sub-images;

associating each of the determined image features with one of the reference features;

forming a frequency distribution representing a frequency of occurrence of the associated reference features for the input digital image;

analyzing the frequency distribution to determine an inequality index, the inequality index being a measure of statistical variation of the reference features;

comparing the inequality index to a predefined threshold to classify the input digital image as a rich-content image or a low-content image;

if the input digital image is classified as a rich-content image:

analyzing the input digital image to automatically determine scene content features relating to spatial structures in the input digital image or semantic content of the input digital image; and

determining the estimated clutter level responsive to the scene content features; and

if the input digital image is classified as a low-content image:

analyzing the input digital image to automatically determine an overall luminance level; and

determining the estimated clutter level responsive to the determined overall luminance level.

2. The method of claim 1 wherein the inequality index is an entropy value.

3. The method of claim 2 wherein the entropy value is determined according to the equation:

E

=

-

∑

k

⁢

p

k

⁢

log

⁡

(

p

k

)

where E is the entropy value, p k is a probability of occurrence of the k th reference feature, and log(•) is a logarithm function.

4. The method of claim 1 wherein the inequality index is a Gini coefficient.

5. The method of claim 4 wherein the Gini coefficient is determined according to the equation:

G

=

∑

k

⁢

(

p

k

)

2

where G is the Gini coefficient, and p k is a probability of occurrence of the k th reference feature.

6. The method of claim 1 wherein the scene content features relating to spatial structures in the input digital image includes a scene asymmetry score providing an indication of a degree of asymmetry of scene content in the input digital image.

7. The method of claim 1 wherein the scene content features relating to spatial structures in the input digital image includes a vanishing point score determined responsive to detecting vanishing points in the input digital image.

8. The method of claim 1 wherein the scene content features relating to semantic content of the input digital image includes a depicted person score determined responsive to depicted person features relating to one or more persons depicted in the input digital image.

9. The method of claim 8 wherein the depicted person features for a depicted person includes a face location, a face size or an identity of the depicted person.

10. The method of claim 1 wherein the scene content features relating to semantic content of the input digital image includes a main subject score determined responsive to one or more main subject features relating to one or more main subjects depicted in the input digital image.

11. The method of claim 10 wherein the main subject features for a main subject include a brightness feature, a contrast feature, an edge distribution feature, a color histogram feature and an inequality index feature, a spatial location feature, a size feature or a subject type feature.

12. The method of claim 1 wherein the reference features and the determined image features are characterized by a set of feature attributes.

13. The method of claim 12 wherein the feature attributes include texture attributes and color attributes.

14. The method of claim 1 wherein a particular image feature is associated with a particular reference feature by:

determining a distance metric between the particular image feature and each of the reference features, and

designating the reference feature having the smallest distance metric to be the associated reference feature.