Method, apparatus and system for detecting a supporting surface region in an image

1. A method of detecting a supporting surface region in an image captured by a camera, said method comprising:

detecting a foreground object in the image;

determining one or more background regions of the image in which a lower part of the detected object exists, based on contact between the one or more background regions and the detected object in the image;

determining a degree of confidence for each of the background regions, the degree of confidence indicating likelihood of a corresponding background region being a supporting surface region, the degree of confidence being based on a number of contacts detected in each region between the foreground object and the one or more background regions;

selecting one or more of the background regions based on each corresponding degree of confidence;

determining similarity of other background regions in the image to at least one of the selected background regions; and

detecting the supporting surface region based on the determined similarity.

2. The method according to claim 1 , wherein the object is detected using statistical background modelling.

3. The method according to claim 1 , wherein background regions are determined using watershed segmentation.

4. The method according to claim 1 , wherein the background regions are determined using geodesic superpixel segmentation.

5. The method according claim 1 , wherein the lower part of the object is determined as a lowest point on the object in the image.

6. The method according to claim 1 , wherein the lower part of the object is a point closest to a central vertical axis of the object on a strongest intensity edge above a lowest point on the object.

7. The method according to claim 1 , wherein the lower part of the object is determined as a point on the object closest to a vertical vanishing point in the image.

8. The method according to claim 1 , further comprising filtering false footprints of a further object in the image, the false footprints being determined according to occlusion of the further object.

9. The method according to claim 1 , further comprising filtering out false footprints based on comparing patches of the image.

10. The method according to claim 1 , wherein false footprints are filtered out using static occlusion likelihood.

11. The method according to claim 1 , wherein the similarity is determined based on colour distribution within each region.

12. The method according to claim 1 wherein the contacts are footprints and the number of contacts detected in each region are detected for the image.

13. The method according to claim 1 wherein the contacts are footprints and the number of contacts detected in each region are detected for a number of further images capturing the one or more regions.

14. An apparatus for detecting a supporting surface region in an image captured by a camera, said apparatus comprising:

a processor; and

a memory storing instructions that, when executed by the processor, cause the apparatus to:

detect a foreground object in the image;

determine one or more background regions of the image in which a lower part of the detected object exists, based on contact between the one or more background regions and the detected object in the image;

determine a degree of confidence for each of the background regions, the degree of confidence indicating likelihood of a corresponding background region being a supporting surface region, the degree of confidence being based on a number of contacts detected in each region between the foreground object and the one or more background regions;

select one or more of the background regions based on each corresponding degree of confidence;

determine similarity of other background regions in the image to at least one of the selected regions; and

detect the supporting surface region based on the determined similarity.

15. A non-transitory computer-readable storage medium having recorded thereon a computer program for detecting a supporting surface region in an image captured by a camera that, said program comprising:

code for detecting a foreground object in the image;

code for determining one or more background regions of the image in which a lower part of the detected object exists, based on contact between the one or more background regions and the detected object in the image;

code for determining a degree of confidence for each of the background regions, the degree of confidence indicating likelihood of a corresponding background region being a supporting surface region, the degree of confidence being based on a number of contacts detected in each region between the foreground object and the one or more background regions;

code for selecting one or more of the background regions based on each corresponding degree of confidence;

code for determining similarity of other background regions in the image to at least one of the selected regions; and

code for detecting the supporting surface region based on the determined similarity.

16. A system for detecting a supporting surface region in an image captured by a camera, said system comprising:

a processor; and

a memory storing instructions that, when executed by the processor, cause the system to:

detect a foreground object in the image;

determine one or more background regions of the image in which a lower part of the detected object exists, based on contact between the one or more background regions and the detected object in the image;

determine a degree of confidence for each of the background regions, the degree of confidence indicating likelihood of a corresponding background region being a supporting surface region, the degree of confidence being based on a number of contacts detected in each region between the foreground object and the one or more background regions;

select one or more of the background regions based on each corresponding degree of confidence;

determine similarity of other background regions in the image to at least one of the selected regions; and

detect the supporting surface region based on the determined similarity.