System, method and apparatus for scene recognition

1. An image processing system for recognizing a scene within an image comprising:

i. an image processing system;

ii. software adapted to operate on the image processing system, the software adapted to generate a distance metric from a set of images, wherein the distance metric indicates a first set of image features and a set of image feature weights corresponding to the first set of image features;

iii. the software further adapted to retrieve an input image;

iv. the software further adapted to extract a set of input image features from the input image, wherein the set of input image features corresponds to the distance metric;

v. the software further adapted to extract a second set of image features from each image in the set of images, wherein each of the extracted sets of image features corresponds to the distance metric;

vi. based on the set of image feature weights of the distance metric, the software further adapted to compute an image feature distance between the set of input image features and each of the extracted sets of image features;

vii. the software further adapted to select, from the set of images, a subset of images based on the computed image feature distances;

viii. the software further adapted to determine a scene type from the scene types of the subset of images; and

ix. the software further adapted to assign the determined scene type to the input image.

2. The image processing system of claim 1 wherein the software application is adapted to retrieve a source scene image via a network interface and segment the source scene image into multiple images, wherein the multiple images include the input image.

3. The image processing system of claim 2 wherein the software application is adapted to:

i. recognize scene types of one or more of the multiple images; and

ii. based on the recognized scene types, determine a scene type for the source scene image.

4. The image processing system of claim 1 wherein the set of input image features includes an estimated depth of the input image and the second set of image features for each image in the set of images includes an estimated depth of the image.

5. The image processing system of claim 1 wherein the software application is adapted to send the distance metric to a client computer.

6. The image processing system of claim 5 wherein the software application is adapted to send the second set of image features for each image in the set of images to the client computer.

7. The image processing system of claim 1 wherein the software application is adapted to receive the second set of image features for the input image from a client computer.

8. The image processing system of claim 1 wherein the image processing system includes an image processing computer and wherein the software application is adapted to retrieve the input image via a network interface coupled to the image processing computer.

9. The image processing system of claim 1 wherein the image processing system is a distributed computing system.

10. The image processing system of claim 9 wherein the image processing system is a cloud computing system.

11. The image processing system of claim 1 wherein the software application is further adapted to:

i. extract a set of raw image features for each image in the set of images; and

ii. reduce dimensionality of the sets of raw image features to derive the first set of image features and the set of image feature weights.

12. The image processing system of claim 11 wherein the software application is further adapted to:

i. perform scene understanding on an unlabeled image to match a scene type to the unlabeled image;

ii. determines that the matched scene type is accurate;

iii. label the unlabeled image with the matched scene type; and

iv. generate a refined image distance metric from the set of images and the newly labeled image.

13. The image processing system of claim 1 wherein the software application is further adapted to:

i. based on the distance metric, learn a classification model from a plurality of images of each scene type in a set of scene types;

ii. apply the learned models to the input image to generate a set of matching scores; and

iii. based on the set of matching scores, select the subset of images.

14. A method for recognizing a scene type of an image, the method operating within an image processing computer and comprising:

i. generating a distance metric from a set of images wherein the set of images are retrieved via the network interface of the image processing computer and the distance metric indicates a first set of image features and a set of image feature weights corresponding to the first set of image features;

ii. retrieving an input image via the network interface;

iii. extracting a set of input image features from the input image, wherein the set of input image features corresponds to the distance metric;

iv. extracting a second set of image features from each image in the set of images, wherein each of the extracted sets of image features corresponds to the distance metric;

v. computing an image feature distance between the set of input image features and each of the extracted sets of image features based on the set of image feature weights;

vi. selecting, from the set of images, a subset of images based on the computed image feature distances;

vii. determining a scene type from the scene types of the subset of images; and

viii. assigning the determined scene type to the input image.

15. The method of claim 14 further comprising:

i. sending the distance metric to a client computer via the network interface; and

ii. sending the second set of image features for each image in the set of images to the client computer via the network interface.

16. The method of claim 14 further comprising receiving the second set of image features for the input image from a client computer via the network interface.

17. The method of claim 14 further comprising:

i. extracting a set of raw image features for each image in the set of images; and

ii. reducing dimensionality of the sets of raw image features to derive the first set of image features and the set of image feature weights.

18. The method of claim 14 further comprising:

i. performing scene understanding on an unlabeled image to match a scene type to the unlabeled image;

ii. determining that the matched scene type is accurate;

iii. labeling the unlabeled image with the matched scene type; and

iv. generating a refined image distance metric from the set of images and the newly labeled image.