Adaptive Scene Dependent Filters In Online Learning Environments

• US 20070147678A1
• Filed: 12/13/2006
• Published: 06/28/2007
• Est. Priority Date: 12/22/2005
• Status: Active Grant

• Current Assignee: Honda Research Institute Europe GmbH (Honda Motor Company)
• Original Assignee: Honda Research Institute Europe GmbH (Honda Motor Company)
• Inventors: Wersing, Heiko, Steil, Jochen, Gotting, Michael
• Granted Patent: US 8,238,650 B2
• Time in Patent Office: Days
• US Class Current: 382/160
• CPC Class Codes: G06V 10/454 Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]

Abstract

A method and a system for determining an object segment in an electronic image. Preferably the method or system is sufficiently fast to allow real-time processing. A method for determining an object segment in an electronic image may comprise the steps of unsupervised learning of a multi-feature segmentation and of forming a relevance map. The method may further comprise the step of estimating the probability of a segment belonging to an object by the overlap of the segment and the relevance map in the electronic image.

Claims

1. A computer based method for determining an object segment of an object represented in an electronic image, comprising the steps of:

unsupervised learning of a multi-feature segmentation; and

forming a relevance map.

2. The method according to claim 1, further comprising the step of:

estimating a probability of a segment belonging to the object by the overlap of the segment and the relevance map in the electronic image.

3. The method of claim 2, wherein the step of unsupervised learning of a multi-feature segmentation comprises the steps of:

forming training data vectors {right arrow over (m)} _(x,y) using basic filter maps (F_i);

obtaining codebook vectors {right arrow over (c)}^j from the training data vectors {right arrow over (m)} _(x,y) using a vector quantization network (VQ);

generating adaptive topographic activation maps (V^j) from the training data vectors {right arrow over (m)} _(x,y) and the codebook vectors {right arrow over (c)}^j ; and

binarising the adaptive topographic activation maps (V^j) to obtain binarised adaptive topographic activation maps (B_i).

4. The method of claim 3, wherein the step of generating adaptive topographic maps utilizes a standard vector quantization network VQ with a fixed number of training steps.

5. The method of claim 3, wherein the training data vectors {right arrow over (m)} (x,y) include the pixel positions (x,y) as features.

6. The method of claim 3, wherein each component of the training data vectors {right arrow over (m)} (x,y) are each normalized by their variance σ

(m_i)².

7. The method of claim 3, wherein each component of the training data vectors is weighted by an additional weighting factor (ξ

^I).

8. The method of claim 7, wherein the additional weighting factor (ξ

^I) is heuristically determined.

9. The method of claim 3, wherein the codebook vectors C^j are obtained by the steps of:

drawing a random (x,y)-position from an image;

generating the vector {right arrow over (m)} (x,y) at this position;

compute the minimal distance (d_min) of m(_x,y) to all codebook vectors C^j in the current codebook; and

assigning a new codebook vector ({right arrow over (c)}^j).

10. The method of claim 9, wherein the new codebook vector ({right arrow over (c)}^j) is equal to {right arrow over (m)} _(x,y) if d_min is larger than a threshold value (d′

) and a new {right arrow over (m)} _(x,y) is drawn otherwise.

11. The method of claim 3, wherein the adaptive topographic activation map (V^j) is scene dependent and is computed as V^j_(x,y)=∥

{right arrow over (m)} <sub>(x,y)−

{right arrow over (c)}</sub>^j∥

².

12. The method of claim 11, wherein the scene dependent adaptive topographic activation map (V^j) is binarised by the formula:

$B_{\left(x,y\right)}^j=\{\begin{array}{cc}1& \mathrm{if}<br><br>\text{\hspace{1em}<br><br>}<br><br>{<br><br>{\stackrel{\to <br><br>}{m}}_{\left(x,y\right)}-{\stackrel{\to <br><br>}{c}}^k<br><br>}^2<<br><br>{<br><br>{\stackrel{\to <br><br>}{m}}_{\left(x,y\right)}-{\stackrel{\to <br><br>}{c}}^j<br><br>}^2,\forall <br><br>k\ne <br><br>j\\ 0& \mathrm{else}\end{array}$

13. The method of to claim 2, wherein a relevance mask (I_REL) is computed as an additive superposition from a center map I_C and a disparity map I_DISP.

14. The method of claim 2, wherein the relevance map is used to find out which combination of the ASDF'"'"'s should be selected.

15. The method according to claim 2, further comprising the step of forming a skin color mask/detecting skin color.

16. The method of claim 15, wherein an adaptive skin color segmentation further excludes skin color regions from a final mask.

17. The method of claim 3, wherein the number of pixels (inPix) of the intersection between a relevance mask based upon the relevance map and the binarised topographic activation map (B_i) and the number of pixels (outPix) of the binarised topographic activation map (B_i) without the relevance mask (I_Rel) are used for selecting the suitable masks.

18. The method of claim 17, wherein a probability of mask (B_i) belonging to the object is estimated by the relative frequency outPix/inPix.

19. The method of claim 18, wherein a mask is included in a final segment mask (I_final), if the relative frequency is larger than a given threshold.

20. The method of claim 19, wherein the final segment mask (I_Final) is computed as the additive superposition of the selected B_i and the skin color pixels are removed from this mask

(I_Final=Σ

_iB_i−

I_skin).

21. A computer software program product embodied on a computer readable medium for performing the method of claim 2 when run on a computing device.