IMAGE PROCESSING

• US 20120008858A1
• Filed: 12/07/2009
• Published: 01/12/2012
• Est. Priority Date: 12/16/2008
• Status: Active Grant

• Current Assignee: Staffordshire University
• Original Assignee: Staffordshire University
• Inventors: Sedky, Mohamed Hamed Ismail, Chibelushi, Claude Chilufya, Moniri, Mansour
• Granted Patent: US 9,047,677 B2
• Time in Patent Office: Days
• US Class Current: 382/155
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/254 involving subtraction of images
• CPC Class Codes: G06T 7/90 Determination of colour characteristics

Abstract

An image segmentation method has a training phase and a segmentation phase. In the training phase, a frame of pixellated data from a camera is processed using information on camera characteristics to render it camera independent. The camera independent data are processed using a chosen value of illuminant spectral characteristics to derive reflectivity data of the items in the image. Pixels of high reflectivity are established. Then, using data from the high reflectivity pixels, the actual illuminant spectral characteristics are established. The illuminant data are then processed to determine information on the illumination of the scene represented by the frame of pixellated data to derive reflectivity data of the scene. The segmentation phase comprises operating on a subsequent frame of pixellated data to render it camera independent and using the determined illumination information to process the camera independent data to determine reflectivity data of the scene to derive a foreground mask.

Claims

1. A device for processing an image, comprising:

processing circuitry configured to receive data of a pixellated current image, and to determine from that image data at least one parameter indicative of the characteristics of the illuminant;

the processing circuitry further configured to process the illuminant data to provide data indicative of at least one parameter indicative of the spectral characteristics of the illumination of the scene shown in the current image, and to process the image data thereby to determine the reflectivity of at least one surface of at least one object shown in the current image;

modelling circuitry configured to derive a background model from information of at least one previous image of the scene; and

decision circuitry configured to compare said reflectivity with a reflectivity of a part of the background model.

2. A device according to claim 1, wherein the processing circuitry is configured to adaptively separate the spectral characteristics of the illumination from the determined surface reflectivity.

3. A device according to claim 1 or 2, wherein the processing circuitry is operable to derive the weights of spectral basis functions indicative of illumination or surface reflectivity for pixels in the current image.

4. A device according to claim 3, wherein the weights are a direct estimate of true parameters of the reflectivity.

5. A device according to any preceding claim, wherein the decision circuitry includes thresholding circuitry whereby only reflectivity differences greater than a threshold are taken into account

6. A device according to any preceding claim, wherein the processing circuitry is configured to normalise the spectral similarity measure to enable selection of pixels indicative of the illumination spectral characteristics.

7. A device according to any preceding claim, further comprising estimating circuitry for estimating the spectral characteristics of the illumination of the scene.

8. A device according to claim 7, wherein the estimating circuitry is operable to determine highlight regions of an input image, and estimate spectral characteristics of the illuminant of the scene therefrom.

9. A device according to claim 7, wherein the estimating circuitry is operable to determine highlight regions of an input image, and estimate the correlated colour temperature of the illuminant of the scene

10. A device according to any preceding claims, further comprising circuitry for determining the spectral characteristics of the illumination using data from the current image and estimating therefrom the spectral characteristics of the illumination.

11. A device according to any preceding claim, wherein the modelling circuitry is operable to derive the background model using statistical parameters of surface reflectivity and illuminant characteristics.

12. A device according to any preceding claim, wherein the modelling circuitry includes update circuitry whereby a change in illuminant triggers an update of model parameters.

13. A device according to any preceding claim, further comprising mask-production circuitry operable to use the result of said decision circuitry to create a segmented mask

14. A device according to claim 13 further comprising a mask processing stage

16. A device according to claim 14 wherein the mask processing stage is configured to apply dilation and erosion processes to the segmented mask

17. A device according to any preceding claim, further comprising a feedback to the modelling circuitry, whereby the background model is updated using information from the current image.

18. Object segmentation apparatus comprising a device according to any preceding claim.

19. A surveillance system comprising object segmentation apparatus of claim 18.

20. A method of processing an image;

the method comprising;

processing pixellated image data to determine from that image data at least one parameter indicative of the characteristics of the illuminant;

processing image data to derive at least one parameter indicative of the spectral characteristics of the illumination of the scene shown in the current image,

processing image data to determine the reflectivity of at least one surface of at least one object shown in the current image; and

comparing said reflectivity with a reflectivity of a part of a background model of the scene shown in the image, the background model being derived from information of at least one previous image of the scene.

21. A method according to claim 20, wherein the processing step comprises deriving the weight of spectral basis functions indicative of surface reflectivity or illumination for pixels in the current image.

22. A method according to claim 20 or 21, wherein the comparing step comprises a threshold comparison.

23. A method according to any of claims 20-22, comprising calculating a ratio between the autocorrelation of the pixels of an image and the autocorrelation of a mean background image.

24. A method according to any of claims 20-23, comprising using the result of said comparing step to create a segmented mask

25. A method according to claim 24, further comprising a mask processing stage

26. A method according to claim 25 wherein the mask processing stage comprises dilation and erosion processes.

27. A method according to any of claims 20-26 further comprising a model maintenance process, whereby the background model is updated using information from the current image.

28. An image segmentation method comprising a training phase, and a segmentation phase, wherein the training phase comprises operating on a frame of pixellated data from a camera using information on camera characteristics to render it camera-independent, and processing the camera independent data using a chosen value of illuminant spectral characteristics to derive reflectivity data whereby pixels of high reflectivity are established, using data from the high reflectivity pixels, determining data on the actual illuminant spectral characteristics and processing the illuminant data to determine information on the illumination of the scene represented by the frame of pixellated data to derive reflectivity data of the scene;

the segmentation phase comprises operating on a subsequent frame of pixellated data to render it camera-independent and using the determined illumination information to process the camera independent data to determine reflectivity data of the scene to derive a foreground mask.