30 views

Uploaded by taulau1234

Các phương pháp dùng phát hiện dáng

save

You are on page 1of 27

**¨ Harri Auvinen, Tapio Leppalampi, Joni Taipale and Maria Teplykh
**

Lappeenranta University of Technology Machine Vision and Digital Image Analysis

November 24th , 2009

Introduction

Scale-Invariant Feature Transform (SIFT) A method developed by David G. Lowe Feature extraction method Invariance in feature extraction A method should locate features Extraction method should be robust, it should handle different types of changes between images

Illumination Afﬁne transform

Scale Rotation

Algorithm

The steps of the SIFT algorithm: 1. Scale-space extrema detection

Search over scales and image locations Locate local extremas

2. Keypoint localization

Selects keypoints from local extremas Keypoints are selected based on measures of their stability

3. Orientation assignment

Orientations are assigned to each keypoint based on local image gradient directions

4. Keypoint descriptor

Local image gradients are measured at the selected scale in the region around each keypoint These are transformed into a representation that allows for signiﬁgant levels of local shape distortion and change in illumination

Scale-space extrema detection

The ﬁrst stage in keypoint detection is to ﬁnd the local extremas in scale-space. It contains A cascade ﬁltering approach Creation of octaves and Scale-space images for each octave

**Scale-space extrema detection
**

The original image I(x, y ) is blurred with Gaussian ﬁlter L(x, y , σ) = G(x, y, σ) ∗ I(x, y ), where G(x, y , σ) = 1 1 exp − 2 (x 2 + y 2 ) . 2πσ 2 2σ (1)

(2)

The procedure is repeated by changing the scale σ by multiplying it with the factor k s times. Then the difference of Gaussians (DoG), D(x, y, σ) = L(x, y , k σ) − L(x, y, σ), (3)

is calculated for adjagent blurred images. According to Lowe to achieve stable keypoints one should set s = 3 and k = 21/s .

Scale-space extrema detection

Scale-space extrema detection

The procedure to calculate the differences of Gaussians is then repeated for each octave. The creation of the next octave: Select the Gaussian blurred image which has σ value twice to that of the original Subsample the image and use the output as the starting point for next octave Subsampling is made by selecting every second pixel from the rows and columns of the image

Keypoint localization

**Detection of local extremas/keypoints
**

Find the extrema points in the DoG pyramid Improve the localization of the keypoint to subpixel accuracy by using a second order Taylor series expansion

Elimination of keypoints

Eliminate some points from the candidate list of keypoints by ﬁnding those that have low contrast or are poorly localised on an edge Contrast thresholding Cornerness thresholding

Detection of local extremas/keypoints

To detect the local maxima and minima of D(x, y , σ) each point is compared with the pixels of all its 26 neighbours If this value is the minimum or maximum, then this point is an extrema

**Detection of local extremas/keypoints. Brown and Lowe method
**

Improvement to matching and stability Approach uses the Taylor expansion of the scale-space function, D(x, y, σ), shifted so that the origin is at the sample point: ∂D T 1 ∂2D D(X ) = D + X+ X X (4) ∂X 2 ∂X 2 where D and its derivatives are evaluated at the sample point and X = (x, y , σ)T is the offset from this point. The location of ˆ the extremum X is determined by taking the derivative of this function with respect to X and setting it to zero, giving ˆ X =− ∂2D ∂X 2

−1

∂D ∂X

(5)

ˆ If X > 0.5 then it means that the extremum lies closer to a different sample point. In this case, the interpolation is performed.

Elimination of keypoints

a The 233x189 pixel original image b The initial 832 keypoints locations at maxima and minima of the difference-of-Gaussian function c After applying a threshold on minimum contrast 729 keypoints remain d The ﬁnal 536 keypoints that remain following an additional threshold on ratio of principal curvatures

Contrast thresholding

ˆ The function value at the extremum, D(X ), is useful for rejecting unstable extrema with low contrast. This can be obtained by substituting equation (5) into (4), giving ˆ D(X ) = D + 1 ∂D T ˆ X. 2 ∂X (6)

ˆ If the function value at X is below a threshold value this point is ˆ excluded. For (c) all extrema with a value of |D(X )| < 0.03 were discarded.

Cornerness thresholding

A poorly deﬁned peak in the difference-of-Gaussian function will have a large principal curvature across the edge but a small one in the perpendicular direction. The principal curvatures can be computed from a 2x2 Hessian matrix, H, computed at the location and scale of the keypoint: H = Dxx Dxy Dxy Dyy (7)

The derivatives are estimated by taking differences of neighboring sample points. The eigenvalues of H are proportional to the principal curvatures of D.

Cornerness thresholding

Let α be the eigenvalue with the largest magnitude and β be the smaller one Tr (H) = Dxx + Dyy = α + β (8) Det(H) = Dxx Dyy − (Dxy )2 = αβ Let r be the ratio between the largest magnitude eigenvalue and the smaller one, so that α = r β. Then, Tr (H)2 (α + β)2 (r β + β)2 (r + 1)2 = = = Det(H) αβ r r β2 The quantity (r + 1)2 /r is at a minimum when the two eigenvalues are equal and it increases with r . Therefore, to check that the ratio of principal curvatures is below some threshold, r , we only need to check Tr (H)2 (r + 1)2 < Det(H) r The transition from (c) to (d) was obtained with r = 10. (11) (10) (9)

Orientation assignment (2)

Left: The point in the middle is the keypoint candidate. The orientations of the points in the square area around this point are precomputed using pixel differences. Right: Each bin in the histogram holds 10 degree, so it covers the whole 360 degree with 36 bins in it. The value of each bin holds the magnitude sums from all the points precomputed within that orientation.

Keypoint descriptor

Keypoint samples are accumulated into orientation histograms summarizing the contents over 4x4 subregions Best result is obtained 4X4 array of histograms with 8 orientation bins in each As a result a 4x4x8 = 128 element feature vector is generated for each keypoint

Keypoint descriptor

Orientation invariance

In order to achieve orientation invariance the coordinates of the descriptor and the gradient orientations are rotated relative to the keypoint orientation For efﬁciency, the gradients are precomputed for all levels of the pyramid A Gaussian weighting function with equal to one half the width of the descriptor window is used to assign a weight to the magnitude of each sample point The purpose of the Gaussian window is

To avoid sudden changes in the descriptor with small changes in the position of the window And to give less emphasis to gradients that are far from the center of the descriptor, as these are most affected by misregistration errors

Boundary affects

To avoid all boundary affects Trilinear interpolation is used to distribute the value of each gradient sample into adjacent histogram bins In other words, each entry into a bin is multiplied by a weight of 1 − d for each dimension d is the distance of the sample from the central value of the bin as measured in units of the histogram bin spacing

Effect of illumination

The feature vector modiﬁcation Reason by this is to reduce the effects of illumination change First, the vector is normalized to unit length Second, threshold the values in the unit feature vector And then renormalizing to unit length

Demo and applications

**Search for the sample in the image
**

Classiﬁcation of remote sensed imagery. [Yang&Newsam, 2008]

Model images of planar objects Recognition of 3D objects Recognising Panoramas People Redetection [Hu et al., 2008]

Object recognition

Recognising panoramas

**Comparison and Modiﬁcations
**

Until now, SIFT has been proven to be the most reliable descriptor among the others. Ancuti&Bekaert, 2007 Mikolajczyk&Schmid, 2005 Modiﬁcations CSIFT: A SIFT Descriptor with Color Invariant Characteristics [Abdel-Hakim&Farag] SIFT-CCH: Increasing the SIFT distinctness by Color Co-occurrence Histograms [Ancuti&Bekaert, 2007] PCA-SIFT: A More Distinctive Representation for Local Image Descriptors [Ke et al., 2004]

”. . . instead of using SIFT’s smoothed weighted histograms, we apply Principal Components Analysis (PCA) to the normalized gradient patch.” ”. . . more distinctive and more compact leading to signiﬁcant improvements in matching accuracy (and speed) for both controlled and real-world conditions.”

References

David G. Lowe, Object Recognition from Local Scale-Invariant Features, Proc. of the International Cenference on Computer Vision, 1999 David G. Lowe, Distinctive Image Features from Scale-Invariant Keypoints, International Journal of Computer Vision, 2004 M. Brown and D.G. Lowe, Recognising Panoramas, International Conference on Computer Vision, 2002 Andrea Vevaldi, SIFT for Matlab, http://www.vlfeat.org/ vedaldi/code/sift.html Cosmin Ancuti and Philippe Bekaert, SIFT-CCH: Increasing the SIFT distinctness by Color Co-occurrence Histograms, Proceedings of 5th IEEE International Symposium on Image and Signal Processing and Analysis, 2007.

References

K. Mikolajczyk and C. Schmid. A performance evaluation of local descriptors. IEEE PAMI, 2005. Alaa E. Abdel-Hakim, Aly A. Farag: CSIFT: A SIFT Descriptor with Color Invariant Characteristics. Y. Ke, R. Suthankar and L. Hutson, PCA-SIFT: a more distinctive representation for local image descriptors, in Proc. of CVPR, 2004. Y. Yang and S. Newsam, Comparing SIFT Descriptors and Gabor Texture Features for Classiﬁcation of Remote Sensed Imagery, IEEE International Conference on Image Processing, 2008 Lei Hu, Shuqiang Jiang, Qingming Huang, Yizhou Wang,Wen Gao,PEOPLE RE-DETECTION USING ADABOOST WITH SIFT AND COLOR CORRELOGRAM, The International Conference on Image Processing (ICIP2008)

- 1-s2.0-S1877705814001350-mainUploaded byAnonymous iTzCnM
- 4 1DoneUploaded byeperla
- Ijarece Vol 3 Issue 9 990 994Uploaded bylakshmipo
- Multiplicadores de Lagrange MatlabUploaded byLeonardo Augusto
- Aplications of Derivatives - ErrorsUploaded bySaahil Ledwani
- bayes basicsUploaded byphli
- Day1Uploaded byidiotfellow
- pknUploaded byfharid
- Appendix B Computer Algebra System B ClassPad 330.pdfUploaded byLi Andrew Marcelo
- Chapter SemanticsExtractionFromImagesUploaded bygangasagarjadhav
- COOptimal Coordination of Directional Overcurrent Relays Using AI Algorithms and ComparisonUploaded byrobertovm2002
- Calculus of VariationsUploaded byPedro Palacios Avila
- optimizationhw2Uploaded bySilvio de Paula
- Fortran IntroUploaded byMathias Eder
- Matlab Plotting lectureUploaded byMister Sinister
- sequences and series notesUploaded byapi-231638444
- Taller PID Control ADRCUploaded byCARLOS EDUARDO HERNANDEZ MEDINA
- 01338475Uploaded byapi-3697505
- 2Uploaded bysalagasim
- New York university mathematical modeling lecture notesUploaded bygohan48
- Ch 2 SolutionsUploaded bymohamed
- CH12RPUploaded byAmar Mokha
- Vitmee11 BrochureUploaded byRakesh Ganwani
- Is 502 Analisis Matematico IIIUploaded byTerry Huaman Che
- hardPROB.pdfUploaded byklausfuchs
- Fuzzy Sets SystemsUploaded byKamleshGolhani
- Chap2Uploaded byIlman Faiq
- 2D Strains and Displacements (1)Uploaded bycoronasean_88804
- P1Uploaded byhltkrgz
- 2009 F4 Mid Firstterm TestUploaded bychiradzulu

- IELTS Advantage Writing.pdfUploaded bytaulau1234
- IELTS Advantage WritingUploaded bytaulau1234
- đình long ielts_Essential Reading for IELTS.pdfUploaded bytaulau1234
- Effective Use CasesUploaded bytaulau1234
- Prefixes and SuffixesUploaded bytaulau1234
- tvs08.intro.pdfUploaded bytaulau1234
- 101 Helpful Hints for IELTS.Uploaded bytaulau1234
- hw1Uploaded bytaulau1234
- hw1Uploaded bytaulau1234

- Appearance Based MethodsUploaded bymastanvalishaik
- Detecting Forgery in Duplicated Region Using Superpixel Segmentation and Feature Point MatchingUploaded byJASH MATHEW
- Optimal Clustering Technique for Handwritten Nandinagari Character RecognitionUploaded byATS
- Face Recognition Using SURF Interest PointsUploaded byaronvirginas
- An Approach for Copy-Move Attack Detection and Transformation RecoveryUploaded byIRJET Journal
- Multi robot exploration and MappingUploaded bymdemirst
- An Adaptive Image-stitching Algorithm for an Underwater Moitoring SystemUploaded bybasela2010
- 3. Overstated Role of CapitalUploaded byjaiswalswatin87
- Blind Assistant _rev 03Uploaded byianni67
- SiftUploaded bytrantinh1010
- Comparison of Feature Detection and Matching Approaches: SIFT and SURFUploaded byGRD Journals
- Tempered Image DetectionUploaded byjkumardk
- study radiometric variationsUploaded byBaga
- Content Based Video Retrieval SystemsUploaded byijujournal
- Google Goggles pptUploaded byLijo James
- LSTM Based Scene LabellingUploaded byAnshuman Sabath
- 074Uploaded bykizzzer
- remotedassensing-04-01392Uploaded byCris Cavali
- report image processingUploaded byMahendra Singh
- Face Recognition Using SIFT- PCA Feature Extraction and SVM ClassifierUploaded byIOSRjournal
- Implementation of Vision based Aero Plane Safe Landing-Area Detection SystemUploaded byInternational Journal for Scientific Research and Development - IJSRD
- 10.1016j.procs.2015.07.547.pdfUploaded byBunda Yosiniella Dianbayu
- A Study of Hand Gesture Recognition TechniqueUploaded byIRJET Journal
- ajest_id7.pdfUploaded byOscar Cobb
- Adaptive Neuro-Fuzzy Inference System (ANFIS) for segmentation of image ROI and Retrieval of ROI based on MP-KDDUploaded byIRJET Journal
- 3D_OdometryUploaded byLegna Lopez
- A Survey On Image Matching MethodsUploaded byInternational Journal of Latest Research in Engineering and Technology
- Human Activity Analysis_ a ReviewUploaded byAlan Mariano Cabello Ortega
- SERRE.2007_Robust object recognition with cortex-like mechanism.pdfUploaded byAnonymous PsEz5kGVae
- Visual Sfm TutorialUploaded byOmar Pech