SlideShare une entreprise Scribd logo
1  sur  30
Outline
 Introduction to SIFT
 Overview of Algorithm
     Construction of Scale space
     DoG (Difference of Gaussian Images)
     Finding Keypoint
     Getting Rid of Bad Keypoint
     Assigning an orientation to keypoints
     Generate SIFT features
Introduction to SIFT
   Scale-invariant feature transform (or SIFT)
    is an algorithm in computer vision to detect
    and describe local features in images.

   This algorithm was published by David
    Lowe.
Types of invariance
   Illumination
Types of invariance
 Illumination
 Scale
Types of invariance
 Illumination
 Scale
 Rotation
Types of invariance
 Illumination
 Scale
 Rotation
 Full perspective
SIFT Algorithm
1. Constructing Scale space
 In scale Space we take the image and
  generate progressively blurred out
  images, then resize the original image to
  half and generate blurred images.
 Images that are of same size but different
  scale are called octaves.
How Blurring is performed?
   Mathematically blurring is defined as convolution of
    Gaussian operator and image.




    where G= Gaussian Blur operator
 L is a blurred image
 G is the Gaussian Blur operator
 I is an image
 x, y are the location coordinates
 σ is the “scale” parameter. The amount of
  blur. Greater the value, greater the blur.
 The * is the convolution operation in x and y.
  It applies Gaussian blur G onto the image I
2. Difference of Gaussian(DoG)
 LoG are obtained by taking second order
  derivative.
 DoG images are equivalent to Laplacian of
  Gaussian image. Moreover DoG are scale
  invariant.
 In other word when we do difference of
  gaussian images, it is multiplied with σ2
  which is present in gaussian blur operator G.
3. Finding Keypoint
   Finding keypoint is a two step process:
    1. Locate maxima/minima in DoG images
    2. Find subpixel maxima/minima
Locate maxima/minima

 In the image X is current pixel, while green
  circles are its neighbors, X is marked as
  Keypoint if it is greatest or east of all 26
  neighboring pixels.
 First and last scale are not checked for
  keypoints as there are not enough neighbors
  to compare.
Find subpixel maxima/minima

 Sub-pixel value are generated using Taylor
  expansion of image around the keypoint found.
 The extreme points of this equation can be found
  by differentiating and equating to zero. On
  solving, we’ll get subpixel key point locations.
  These subpixel values increase chances of
  matching and stability of the algorithm.
4. Eliminating bad keypoints
1.       Removing Low Contrast features
      If magnitude of intensity at current pixel is less
         than certain value then it is rejected.
2.       Removing edges
         For poorly defined peaks in the DoG
          function, the principal curvature across the edge
          would be much larger than the principal
          curvature along it
         To determine edges Hessian matrix is used.
Tr (H) = Dxx + Dyy
Det(H) = DxxDyy - (Dxy )2



   If the value of R is greater for a candidate
    keypoint, then that keypoint is poorly localized and
    hence rejected.
5. Assigning Orientation
 Gradient direction and magnitude around
  keypoints are collected, and prominent
  orientations are assigned to keypoints.
 Calculations are done relative to this
  orientation, hence it ensure rotation
  invariance.
   The magnitude and orientation is calculated for
    all pixels around the keypoint.
    Then, A histogram is created for this.




   So, orientation can split up one keypoint into
    multiple keypoints
6. Generating SIFT Features
  Creating fingerprint for each keypoint, so that
   we can distinguish between different keypoints.
  A 16 x 16 window is taken around
   keypoint, and it is divided into 16 4 x 4
   windows.
Generating SIFT Features
   Within each 4×4 window, gradient magnitudes
    and orientations are calculated. These
    orientations are put into an 8
    bin histogram, depending on gradient directions.
Generating SIFT Features
 The value added to bin also depend upon distance
  from keypoint ,so gradients which are far are less in
  magnitude.
 This is achieved by using Gaussian weighting
  function.
   This has to be repeated for all 16 4x4 regions so
    we will get total 16x8=128 numbers. These 128
    numbers are normalized and resultant 128 numbers
    form feature vector which determine a keypoint
    uniquely.
Problem associated with feature vector
1. Rotation Dependence
   If we rotate the image all the gradient orientation will
    get change. So to avoid this keypoint’s rotation is
    subtracted from each gradient orientation. Hence
    each gradient orientation is relative to keypoint’s
    orientation.
 2. Illumination Dependence
   If we threshold numbers that are big, we can achieve
    illumination independence. So, any number (of the
    128) greater than 0.2 is changed to 0.2. This resultant
    feature vector is normalized again. And now we have
    an illumination independent feature vector.
Application
   Application of SIFT include object
    recognition, gesture recognition, image
    stitching, 3D modeling.
Object recognition
Image
Stitching
References
   http://www.aishack.in/2010/05/sift-scale-invariant-
    feature-transform
   http://en.wikipedia.org/wiki/Scale-invariant feature
    transform
   yumeng-SIFTreport-5.18_bpt.pdf
   Paper on SIFT by Harri Auvinen, Tapio
    Lepp¨alampi, Joni Taipale and Maria Teplykh.
   David G. Lowe, Distinctive Image Features from
    Scale-Invariant Keypoints, International Journal of
    Computer Vision, 2004

Contenu connexe

Tendances

Image segmentation ppt
Image segmentation pptImage segmentation ppt
Image segmentation ppt
Gichelle Amon
 

Tendances (20)

Chapter10 image segmentation
Chapter10 image segmentationChapter10 image segmentation
Chapter10 image segmentation
 
Color Image Processing
Color Image ProcessingColor Image Processing
Color Image Processing
 
Feature detection and matching
Feature detection and matchingFeature detection and matching
Feature detection and matching
 
Presentation of Lossy compression
Presentation of Lossy compressionPresentation of Lossy compression
Presentation of Lossy compression
 
Edge linking in image processing
Edge linking in image processingEdge linking in image processing
Edge linking in image processing
 
Edge Detection and Segmentation
Edge Detection and SegmentationEdge Detection and Segmentation
Edge Detection and Segmentation
 
Image segmentation
Image segmentation Image segmentation
Image segmentation
 
Hog and sift
Hog and siftHog and sift
Hog and sift
 
Spatial Filters (Digital Image Processing)
Spatial Filters (Digital Image Processing)Spatial Filters (Digital Image Processing)
Spatial Filters (Digital Image Processing)
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
Image Filtering in the Frequency Domain
Image Filtering in the Frequency DomainImage Filtering in the Frequency Domain
Image Filtering in the Frequency Domain
 
Introduction to object detection
Introduction to object detectionIntroduction to object detection
Introduction to object detection
 
Fields of digital image processing slides
Fields of digital image processing slidesFields of digital image processing slides
Fields of digital image processing slides
 
Wavelet transform in image compression
Wavelet transform in image compressionWavelet transform in image compression
Wavelet transform in image compression
 
Image segmentation
Image segmentationImage segmentation
Image segmentation
 
Image segmentation ppt
Image segmentation pptImage segmentation ppt
Image segmentation ppt
 
Edge detection
Edge detectionEdge detection
Edge detection
 
Image Enhancement in Spatial Domain
Image Enhancement in Spatial DomainImage Enhancement in Spatial Domain
Image Enhancement in Spatial Domain
 
Lecture 1 for Digital Image Processing (2nd Edition)
Lecture 1 for Digital Image Processing (2nd Edition)Lecture 1 for Digital Image Processing (2nd Edition)
Lecture 1 for Digital Image Processing (2nd Edition)
 
Digital Image Processing
Digital Image ProcessingDigital Image Processing
Digital Image Processing
 

En vedette

An improved double coding local binary pattern algorithm for face recognition
An improved double coding local binary pattern algorithm for face recognitionAn improved double coding local binary pattern algorithm for face recognition
An improved double coding local binary pattern algorithm for face recognition
eSAT Journals
 
Towards Efficient Privacy-preserving Image Feature Extraction in Cloud Computing
Towards Efficient Privacy-preserving Image Feature Extraction in Cloud ComputingTowards Efficient Privacy-preserving Image Feature Extraction in Cloud Computing
Towards Efficient Privacy-preserving Image Feature Extraction in Cloud Computing
Si Chen
 
Tính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộc
Tính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộcTính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộc
Tính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộc
Chien Dang
 
Modern features-part-2-descriptors
Modern features-part-2-descriptorsModern features-part-2-descriptors
Modern features-part-2-descriptors
zukun
 
Contextless Object Recognition with Shape-enriched SIFT and Bags of Features
Contextless Object Recognition with Shape-enriched SIFT and Bags of FeaturesContextless Object Recognition with Shape-enriched SIFT and Bags of Features
Contextless Object Recognition with Shape-enriched SIFT and Bags of Features
Universitat Politècnica de Catalunya
 

En vedette (20)

Scale Invariant feature transform
Scale Invariant feature transformScale Invariant feature transform
Scale Invariant feature transform
 
Feature Matching using SIFT algorithm
Feature Matching using SIFT algorithmFeature Matching using SIFT algorithm
Feature Matching using SIFT algorithm
 
SIFT Algorithm Introduction
SIFT Algorithm IntroductionSIFT Algorithm Introduction
SIFT Algorithm Introduction
 
Scale Invariant Feature Tranform
Scale Invariant Feature TranformScale Invariant Feature Tranform
Scale Invariant Feature Tranform
 
Scale invariant feature transform
Scale invariant feature transformScale invariant feature transform
Scale invariant feature transform
 
MIT 6.870 - Template Matching and Histograms (Nicolas Pinto, MIT)
MIT 6.870 - Template Matching and Histograms (Nicolas Pinto, MIT)MIT 6.870 - Template Matching and Histograms (Nicolas Pinto, MIT)
MIT 6.870 - Template Matching and Histograms (Nicolas Pinto, MIT)
 
Local binary pattern
Local binary patternLocal binary pattern
Local binary pattern
 
Feature Extraction
Feature ExtractionFeature Extraction
Feature Extraction
 
face recognition system using LBP
face recognition system using LBPface recognition system using LBP
face recognition system using LBP
 
An improved double coding local binary pattern algorithm for face recognition
An improved double coding local binary pattern algorithm for face recognitionAn improved double coding local binary pattern algorithm for face recognition
An improved double coding local binary pattern algorithm for face recognition
 
Face Recognition Techniques
Face Recognition TechniquesFace Recognition Techniques
Face Recognition Techniques
 
Edge detection-LOG
Edge detection-LOGEdge detection-LOG
Edge detection-LOG
 
Towards Efficient Privacy-preserving Image Feature Extraction in Cloud Computing
Towards Efficient Privacy-preserving Image Feature Extraction in Cloud ComputingTowards Efficient Privacy-preserving Image Feature Extraction in Cloud Computing
Towards Efficient Privacy-preserving Image Feature Extraction in Cloud Computing
 
Preprocessing
PreprocessingPreprocessing
Preprocessing
 
Fuzzy Logic Based Edge Detection
Fuzzy Logic Based Edge DetectionFuzzy Logic Based Edge Detection
Fuzzy Logic Based Edge Detection
 
Dsip and its biometrics appln
Dsip and its biometrics applnDsip and its biometrics appln
Dsip and its biometrics appln
 
Tính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộc
Tính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộcTính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộc
Tính toán khoa học - Chương 7: Các phương pháp cực tiểu hóa không ràng buộc
 
Modern features-part-2-descriptors
Modern features-part-2-descriptorsModern features-part-2-descriptors
Modern features-part-2-descriptors
 
Lec04 gpu architecture
Lec04 gpu architectureLec04 gpu architecture
Lec04 gpu architecture
 
Contextless Object Recognition with Shape-enriched SIFT and Bags of Features
Contextless Object Recognition with Shape-enriched SIFT and Bags of FeaturesContextless Object Recognition with Shape-enriched SIFT and Bags of Features
Contextless Object Recognition with Shape-enriched SIFT and Bags of Features
 

Similaire à SIFT

An automatic algorithm for object recognition and detection based on asift ke...
An automatic algorithm for object recognition and detection based on asift ke...An automatic algorithm for object recognition and detection based on asift ke...
An automatic algorithm for object recognition and detection based on asift ke...
Kunal Kishor Nirala
 

Similaire à SIFT (20)

Sift detector boosted by adaptive contrast threshold to improve matching robu...
Sift detector boosted by adaptive contrast threshold to improve matching robu...Sift detector boosted by adaptive contrast threshold to improve matching robu...
Sift detector boosted by adaptive contrast threshold to improve matching robu...
 
Sift detector boosted by adaptive contrast threshold to improve matching robu...
Sift detector boosted by adaptive contrast threshold to improve matching robu...Sift detector boosted by adaptive contrast threshold to improve matching robu...
Sift detector boosted by adaptive contrast threshold to improve matching robu...
 
Literature Survey on Interest Points based Watermarking
Literature Survey on Interest Points based WatermarkingLiterature Survey on Interest Points based Watermarking
Literature Survey on Interest Points based Watermarking
 
CV PPT.pptx
CV PPT.pptxCV PPT.pptx
CV PPT.pptx
 
Ijcatr04041016
Ijcatr04041016Ijcatr04041016
Ijcatr04041016
 
06 image features
06 image features06 image features
06 image features
 
SIFT.ppt
SIFT.pptSIFT.ppt
SIFT.ppt
 
An automatic algorithm for object recognition and detection based on asift ke...
An automatic algorithm for object recognition and detection based on asift ke...An automatic algorithm for object recognition and detection based on asift ke...
An automatic algorithm for object recognition and detection based on asift ke...
 
Intelligent Auto Horn System Using Artificial Intelligence
Intelligent Auto Horn System Using Artificial IntelligenceIntelligent Auto Horn System Using Artificial Intelligence
Intelligent Auto Horn System Using Artificial Intelligence
 
EFFECTIVE INTEREST REGION ESTIMATION MODEL TO REPRESENT CORNERS FOR IMAGE
EFFECTIVE INTEREST REGION ESTIMATION MODEL TO REPRESENT CORNERS FOR IMAGE EFFECTIVE INTEREST REGION ESTIMATION MODEL TO REPRESENT CORNERS FOR IMAGE
EFFECTIVE INTEREST REGION ESTIMATION MODEL TO REPRESENT CORNERS FOR IMAGE
 
image segmentation by ppres.pptx
image segmentation by ppres.pptximage segmentation by ppres.pptx
image segmentation by ppres.pptx
 
GEOMETRIC TAMPERING ESTIMATION BY MEANS OF A SIFT-BASED FORENSIC ANALYSIS
GEOMETRIC TAMPERING ESTIMATION  BY MEANS OF A SIFT-BASED FORENSIC ANALYSISGEOMETRIC TAMPERING ESTIMATION  BY MEANS OF A SIFT-BASED FORENSIC ANALYSIS
GEOMETRIC TAMPERING ESTIMATION BY MEANS OF A SIFT-BASED FORENSIC ANALYSIS
 
Improved Characters Feature Extraction and Matching Algorithm Based on SIFT
Improved Characters Feature Extraction and Matching Algorithm Based on SIFTImproved Characters Feature Extraction and Matching Algorithm Based on SIFT
Improved Characters Feature Extraction and Matching Algorithm Based on SIFT
 
Practical Digital Image Processing 3
 Practical Digital Image Processing 3 Practical Digital Image Processing 3
Practical Digital Image Processing 3
 
11 cie552 image_featuresii_sift
11 cie552 image_featuresii_sift11 cie552 image_featuresii_sift
11 cie552 image_featuresii_sift
 
Edge Drawing - An Heuristic Approach to Robust Real-Time Edge Detection
Edge Drawing - An Heuristic Approach to Robust Real-Time Edge DetectionEdge Drawing - An Heuristic Approach to Robust Real-Time Edge Detection
Edge Drawing - An Heuristic Approach to Robust Real-Time Edge Detection
 
Estrazione automatica delle linee in un'immagine digitale
Estrazione automatica delle linee in un'immagine digitaleEstrazione automatica delle linee in un'immagine digitale
Estrazione automatica delle linee in un'immagine digitale
 
Practical Digital Image Processing 4
Practical Digital Image Processing 4Practical Digital Image Processing 4
Practical Digital Image Processing 4
 
Practical Digital Image Processing 2
Practical Digital Image Processing 2Practical Digital Image Processing 2
Practical Digital Image Processing 2
 
Lec07 corner blob
Lec07 corner blobLec07 corner blob
Lec07 corner blob
 

Dernier

+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
?#DUbAI#??##{{(☎️+971_581248768%)**%*]'#abortion pills for sale in dubai@
 

Dernier (20)

AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024AXA XL - Insurer Innovation Award Americas 2024
AXA XL - Insurer Innovation Award Americas 2024
 
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemkeProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
ProductAnonymous-April2024-WinProductDiscovery-MelissaKlemke
 
Corporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptxCorporate and higher education May webinar.pptx
Corporate and higher education May webinar.pptx
 
Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)Powerful Google developer tools for immediate impact! (2023-24 C)
Powerful Google developer tools for immediate impact! (2023-24 C)
 
How to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected WorkerHow to Troubleshoot Apps for the Modern Connected Worker
How to Troubleshoot Apps for the Modern Connected Worker
 
Strategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a FresherStrategies for Landing an Oracle DBA Job as a Fresher
Strategies for Landing an Oracle DBA Job as a Fresher
 
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data DiscoveryTrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
TrustArc Webinar - Unlock the Power of AI-Driven Data Discovery
 
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ..."I see eyes in my soup": How Delivery Hero implemented the safety system for ...
"I see eyes in my soup": How Delivery Hero implemented the safety system for ...
 
Ransomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdfRansomware_Q4_2023. The report. [EN].pdf
Ransomware_Q4_2023. The report. [EN].pdf
 
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
Mastering MySQL Database Architecture: Deep Dive into MySQL Shell and MySQL R...
 
ICT role in 21st century education and its challenges
ICT role in 21st century education and its challengesICT role in 21st century education and its challenges
ICT role in 21st century education and its challenges
 
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
+971581248768>> SAFE AND ORIGINAL ABORTION PILLS FOR SALE IN DUBAI AND ABUDHA...
 
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
Apidays New York 2024 - Accelerating FinTech Innovation by Vasa Krishnan, Fin...
 
MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024MINDCTI Revenue Release Quarter One 2024
MINDCTI Revenue Release Quarter One 2024
 
Apidays Singapore 2024 - Modernizing Securities Finance by Madhu Subbu
Apidays Singapore 2024 - Modernizing Securities Finance by Madhu SubbuApidays Singapore 2024 - Modernizing Securities Finance by Madhu Subbu
Apidays Singapore 2024 - Modernizing Securities Finance by Madhu Subbu
 
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost SavingRepurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
Repurposing LNG terminals for Hydrogen Ammonia: Feasibility and Cost Saving
 
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
Emergent Methods: Multi-lingual narrative tracking in the news - real-time ex...
 
A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?A Year of the Servo Reboot: Where Are We Now?
A Year of the Servo Reboot: Where Are We Now?
 
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law DevelopmentsTrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
TrustArc Webinar - Stay Ahead of US State Data Privacy Law Developments
 
Data Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt RobisonData Cloud, More than a CDP by Matt Robison
Data Cloud, More than a CDP by Matt Robison
 

SIFT

  • 1.
  • 2. Outline  Introduction to SIFT  Overview of Algorithm  Construction of Scale space  DoG (Difference of Gaussian Images)  Finding Keypoint  Getting Rid of Bad Keypoint  Assigning an orientation to keypoints  Generate SIFT features
  • 3. Introduction to SIFT  Scale-invariant feature transform (or SIFT) is an algorithm in computer vision to detect and describe local features in images.  This algorithm was published by David Lowe.
  • 4. Types of invariance  Illumination
  • 5. Types of invariance  Illumination  Scale
  • 6. Types of invariance  Illumination  Scale  Rotation
  • 7. Types of invariance  Illumination  Scale  Rotation  Full perspective
  • 9. 1. Constructing Scale space  In scale Space we take the image and generate progressively blurred out images, then resize the original image to half and generate blurred images.  Images that are of same size but different scale are called octaves.
  • 10. How Blurring is performed?  Mathematically blurring is defined as convolution of Gaussian operator and image. where G= Gaussian Blur operator
  • 11.  L is a blurred image  G is the Gaussian Blur operator  I is an image  x, y are the location coordinates  σ is the “scale” parameter. The amount of blur. Greater the value, greater the blur.  The * is the convolution operation in x and y. It applies Gaussian blur G onto the image I
  • 12. 2. Difference of Gaussian(DoG)
  • 13.  LoG are obtained by taking second order derivative.  DoG images are equivalent to Laplacian of Gaussian image. Moreover DoG are scale invariant.  In other word when we do difference of gaussian images, it is multiplied with σ2 which is present in gaussian blur operator G.
  • 14. 3. Finding Keypoint  Finding keypoint is a two step process: 1. Locate maxima/minima in DoG images 2. Find subpixel maxima/minima
  • 15. Locate maxima/minima  In the image X is current pixel, while green circles are its neighbors, X is marked as Keypoint if it is greatest or east of all 26 neighboring pixels.  First and last scale are not checked for keypoints as there are not enough neighbors to compare.
  • 16. Find subpixel maxima/minima  Sub-pixel value are generated using Taylor expansion of image around the keypoint found.  The extreme points of this equation can be found by differentiating and equating to zero. On solving, we’ll get subpixel key point locations. These subpixel values increase chances of matching and stability of the algorithm.
  • 17. 4. Eliminating bad keypoints 1. Removing Low Contrast features  If magnitude of intensity at current pixel is less than certain value then it is rejected. 2. Removing edges  For poorly defined peaks in the DoG function, the principal curvature across the edge would be much larger than the principal curvature along it  To determine edges Hessian matrix is used.
  • 18. Tr (H) = Dxx + Dyy Det(H) = DxxDyy - (Dxy )2  If the value of R is greater for a candidate keypoint, then that keypoint is poorly localized and hence rejected.
  • 19. 5. Assigning Orientation  Gradient direction and magnitude around keypoints are collected, and prominent orientations are assigned to keypoints.  Calculations are done relative to this orientation, hence it ensure rotation invariance.
  • 20. The magnitude and orientation is calculated for all pixels around the keypoint. Then, A histogram is created for this.  So, orientation can split up one keypoint into multiple keypoints
  • 21. 6. Generating SIFT Features  Creating fingerprint for each keypoint, so that we can distinguish between different keypoints.  A 16 x 16 window is taken around keypoint, and it is divided into 16 4 x 4 windows.
  • 22. Generating SIFT Features  Within each 4×4 window, gradient magnitudes and orientations are calculated. These orientations are put into an 8 bin histogram, depending on gradient directions.
  • 23. Generating SIFT Features  The value added to bin also depend upon distance from keypoint ,so gradients which are far are less in magnitude.  This is achieved by using Gaussian weighting function.
  • 24.
  • 25. This has to be repeated for all 16 4x4 regions so we will get total 16x8=128 numbers. These 128 numbers are normalized and resultant 128 numbers form feature vector which determine a keypoint uniquely.
  • 26. Problem associated with feature vector 1. Rotation Dependence  If we rotate the image all the gradient orientation will get change. So to avoid this keypoint’s rotation is subtracted from each gradient orientation. Hence each gradient orientation is relative to keypoint’s orientation. 2. Illumination Dependence  If we threshold numbers that are big, we can achieve illumination independence. So, any number (of the 128) greater than 0.2 is changed to 0.2. This resultant feature vector is normalized again. And now we have an illumination independent feature vector.
  • 27. Application  Application of SIFT include object recognition, gesture recognition, image stitching, 3D modeling.
  • 30. References  http://www.aishack.in/2010/05/sift-scale-invariant- feature-transform  http://en.wikipedia.org/wiki/Scale-invariant feature transform  yumeng-SIFTreport-5.18_bpt.pdf  Paper on SIFT by Harri Auvinen, Tapio Lepp¨alampi, Joni Taipale and Maria Teplykh.  David G. Lowe, Distinctive Image Features from Scale-Invariant Keypoints, International Journal of Computer Vision, 2004