1 / 21

Automated Fingertip Detection

Automated Fingertip Detection. Thesis Defense Presentation by: Joseph Butler. Outline. Introduction Related Work Our solution Color and texture masking Auto-rotation Orientation estimation Poincare index Support v ector classification Connected neighbors and automated cropping

danton
Download Presentation

Automated Fingertip Detection

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Automated Fingertip Detection Thesis Defense Presentation by: Joseph Butler

  2. Outline • Introduction • Related Work • Our solution • Color and texture masking • Auto-rotation • Orientation estimation • Poincare index • Support vector classification • Connected neighbors and automated cropping • Results • Conclusion

  3. Introduction • Fingerprint modality one of the oldest biometric modalities • Extraction has gone from ink to touch sensors and now into digital images • Current work in digital image collection focuses on extraction • Complete automated system includes fingertip detection and extraction

  4. Related Work • Hong et al. use ridge orientation and frequency analysis to generate block specific Gabor filters to further enhance the contrast between friction ridges and valleys • Wang and Wangalso use ridge orientation estimation to calculate Poincare index values per block which are used to locate core and delta points • Lee et al. use a combination color and texture mask to isolate a single fingertip in a digital image • Hiew et al. captured digital images but used a highly controlled capture scenario which left the single preprocessing step of removing a set background color. Once captured the images were enhanced using a Short Time Fourier Transform

  5. Color Mask • Gathered skin-color samples from palms and/or fingers • Convert to Y’UV color space • Samples used to find distribution of U and V • Gaussian bimodal curve best fit for our distributions • Use optimal threshold technique to find threshold between curves

  6. Color Mask • Difference between two curves • Generate binary mask

  7. Texture Mask • Short depth of field given necessity to capture fine detail • Discrete wavelet transform • Two dimensional Haar wavelet • Binary mask • Combine color and texture mask

  8. Auto-rotation • Unrestrained capture • Leverage color mask • Find concentration of unmasked area • Rotate image so concentrated area is at the bottom

  9. Orientation Estimation • Use standard block size as a starting point • Find gradient in X direction and gradient in Y direction • Compute gradient average of entire block

  10. Orientation Estimation • Find ridge width using gradient average value • Resize blocks based on ridge width • Recalculate gradient average • Orthogonal to gradient is ridge orientation

  11. Poincare Index • Leverage orientation of each block k’=(k+1)mod(N)

  12. Poincare Index • A measure of the difference between a block’s orientation value and those of its neighbors Core Delta Delta Core & delta pair

  13. Support Vector Classification • Use training images to classify blocks as core or non-core • Create feature vectors using Poincare values of a block and its neighbors • Cast these feature vectors into a higher dimensional space find best fitting plane that divides the two classes

  14. Support Vector Classification • Classify test image blocks as core or non-core • Differentiate erroneous classifications • True core blocks found in groups

  15. Connected Neighbors and Automated Cropping • Recursively count number of connected neighbors • Identify core region

  16. Results • Our collection • Web collection • Number of fingertips that are identifiable • Positive detection rate • Expected versus actual

  17. Example from web collection

  18. Good Example from our collection

  19. BadExample from our collection

  20. Conclusion • Web collection had positive detection rate of 67.83% • Our collection had positive detection rate of 68.75% • Uncontrolled capture is difficult • Room for improvement • Future work

  21. References • C.C. Chang and C.J. Lin. LIBSVM: a library for support vector machines. ACM Transactions on Intelligent Systems and Technology, pages 27:1{27:27, 2011. Software available at http://www.csie.ntu.edu.tw/cjlin/libsvm. • B.Y. Hiew, A.B.J. Teoh, and D.C.L. Ngo. Automatic digital camera based fingerprint image preprocessing. In Proceedings of the IEEE International Conference on Computer Graphics, Imaging and Visualization, pages 182-189, 2006. • C. Lee, S. Lee, J. Kim, and S.J. Kim. Preprocessing of a fingerprint image captured with a mobile camera. In Proceedings of International Conference on Advances in Biometrics, pages 348-355, 2006. • S. Wang and Y. Wang. Fingerprint enhancement in the singular point area. IEEE Signal Processing Letters, 11(1):16 - 19, 2004. • P. Yu, D. Xu, H. Li, and H. Zhou. Fingerprint image preprocessing based on whole-hand image captured by digital camera. In Proceedings of International Conference on Computational Intelligence and Software Engineering, pages 1-4, 2009.

More Related