Method and apparatus for image analysis of a gabor-wavelet transformed image using a neural network

1. A method for classifying a feature in an image frame, comprising:

transforming an original image frame having an array of pixels using wavelet transformations to generate a transformed image frame having an array of pixels, each pixel of the transformed image being associated with a respective pixel of the original image frame and being represented by a predetermined number of wavelet component values;

selecting a pixel of the transformed image frame associated with the feature;

providing a neural network having an output and a predetermined number of inputs, each input being associated with a respective wavelet component value of the predetermined number of wavelet component values of the selected pixel, wherein the neural network is trained to classify the feature in a transformed image frame; and

classifying, using the neural network, the local feature based on the wavelet component values provided at the neural network inputs, and indicating a class of the feature at the neural network output.

2. A method for classifying a feature in an image frame as defined in claim 1 , wherein the wavelet transformations use Gabor wavelets.

3. A method for classifying a feature in an image frame as defined in claim 1 , wherein each wavelet component value is generated based on a Gabor wavelet having a particular orientation and frequency.

4. A method for classifying a feature in an image frame as defined in claim 1 , wherein the feature in the image frame is a facial feature.

5. A method for classifying a feature in an image frame as defined in claim 1 , wherein the predetermined number of wavelet component values and the predetermined number of neural network inputs is 12.

6. A method for classifying a feature in an image frame as defined in claim 1 , wherein the wavelet component values are magnitudes of complex numbers.

7. A method for analyzing an object image in an image frame, comprising:

transforming an original image frame having an array of pixels using wavelet transformations to generate a transformed image frame having an array of pixels, each pixel of the transformed image being associated with a respective pixel of the original image frame and being represented by a predetermined number of wavelet component values;

selecting pixels of the transformed image frame corresponding to sensing nodes of a label graph, each sensing node for analyzing a local feature of the object image in the original image frame;

providing a plurality of neural networks, each neural network being associated with a sensing node of the label graph and having an output and a predetermined number of inputs, each input being associated with a respective wavelet component value of the predetermined number of wavelet component values of the respective selected pixel, wherein each neural network is trained to analyze the respective local feature of the object image in a transformed image frame;

analyzing, using the neural networks, the local features based on the wavelet component values provided at the neural network inputs, and indicating a characteristic of the object image based on the neural network outputs.

8. A method for classifying a feature in an image frame as defined in claim 7 , wherein the wavelet transformations use Gabor wavelets.

9. A method for classifying a feature in an image frame as defined in claim 7 , wherein each wavelet component value is generated based on a Gabor wavelet having a particular orientation and frequency.

10. A method for classifying a feature in an image frame as defined in claim 7 , wherein the object image is a facial image.

11. A method for classifying a feature in an image frame as defined in claim 7 , wherein the predetermined number of wavelet component values is 40 and the predetermined number of neural network inputs is 12.

12. A method for classifying a feature in an image frame as defined in claim 7 , wherein the wavelet component values are magnitudes of complex numbers.

13. Apparatus for classifying a feature in an image frame, comprising:

means for transforming an original image frame having an array of pixels using wavelet transformations to generate a transformed image frame having an array of pixels, each pixel of the transformed image being associated with a respective pixel of the original image frame and being represented by a predetermined number of wavelet component values;

means for selecting a pixel of the transformed image frame associated with the feature;

a neural network having an output and a predetermined number of inputs, each input being associated with a respective wavelet component value of the predetermined number of wavelet component values of the selected pixel, wherein the neural network is trained to classify the feature in a transformed image frame and classifies the local feature based on the wavelet component values provided at the neural network inputs, and indicating a class of the feature at the neural network output.

14. Apparatus for classifying a feature in an image frame as defined in claim 13 , wherein the wavelet transformations use Gabor wavelets.

15. Apparatus for classifying a feature in an image frame as defined in claim 13 , wherein each wavelet component value is generated based on a Gabor wavelet having a particular orientation and frequency.

16. Apparatus for classifying a feature in an image frame as defined in claim 13 , wherein the feature in the image frame is a facial feature.

17. Apparatus for classifying a feature in an image frame as defined in claim 13 , wherein the predetermined number of wavelet component values and the predetermined number of neural network inputs is 12.

18. Apparatus for classifying a feature in an image frame as defined in claim 13 , wherein the wavelet component values are magnitudes of complex numbers.

19. Apparatus for analyzing an object image in an image frame, comprising:

means for transforming an original image frame having an array of pixels using wavelet transformations to generate a transformed image frame having an array of pixels, each pixel of the transformed image being associated with a respective pixel of the original image frame and being represented by a predetermined number of wavelet component values;

means for selecting pixels of the transformed image frame corresponding to sensing nodes of a label graph, each sensing node for analyzing a local features of the object image in the original image frame;

a plurality of neural networks, each neural network being associated with a sensing node of the label graph and having an output and a predetermined number of inputs, each input associated with a respective wavelet component value of the predetermined number of wavelet component values of the respective selected pixel, wherein each neural network is trained to analyze the respective local feature of the object image in a transformed image frame;

means for analyzing, using the neural networks, the local features based on the wavelet component values provided at the neural network inputs, and indicating a characteristic of the object image based on the neural network outputs.

20. Apparatus for classifying a feature in an image frame as defined in claim 19 , wherein the wavelet transformations use Gabor wavelets.

21. Apparatus for classifying a feature in an image frame as defined in claim 19 , wherein each wavelet component value is generated based on a Gabor wavelet having a particular orientation and frequency.

22. Apparatus for classifying a feature in an image frame as defined in claim 19 , wherein the object image is a facial image.

23. Apparatus for classifying a feature in an image frame as defined in claim 19 , wherein the predetermined number of wavelet component values is 40 and the predetermined number of neural network inputs is 40.

24. Apparatus for classifying a feature in an image frame as defined in claim 19 , wherein the wavelet component values are complex numbers.