PIP - Perspective of Image Processing Introduction SIFT - PowerPoint PPT Presentation

PIP - Perspective of Image Processing Introduction SIFT Implementation of Code Rishidhar Reddy Bommu, Bryan Humphreys, Aerin Kim, Neural Networks Sombo Koo K-Means Team Leader: Dhatri Chennavajula Color Cluster Project Director: Roger Dodd CAMCOS Director: Slobodan Simi´ c San Jos´ e State University May 14th, 2015

Outline Introduction Outline Image Processing Edges Feature Introduction to Image Processing Detection SIFT Implementation of Code Neural Networks K-Means Color Cluster

Outline Introduction Outline Image Processing Edges Feature Introduction to Image Processing Detection SIFT The Scale - Invariant Feature Transform (SIFT) Implementation of Code Neural Networks K-Means Color Cluster

Outline Introduction Outline Image Processing Edges Feature Introduction to Image Processing Detection SIFT The Scale - Invariant Feature Transform (SIFT) Implementation Applications of Feature Description of Code Neural Networks K-Means Color Cluster

Outline Introduction Outline Image Processing Edges Feature Introduction to Image Processing Detection SIFT The Scale - Invariant Feature Transform (SIFT) Implementation Applications of Feature Description of Code Neural Future Work Networks K-Means Color Cluster

Introduction to Image Processing Introduction Q: What is an Image? Outline Image Processing Edges Feature Detection SIFT Implementation of Code Neural Networks K-Means Color Cluster

Introduction to Image Processing Introduction Q: What is an Image? Outline Image Processing Edges Feature Detection  11 11 12 12 23  SIFT 12 14 23 22 21   Implementation   of Code − → 22 20 23 24 25   Neural   33 34 36 33 33 Networks   32 31 10 10 15 K-Means Color Cluster A: It’s an array of numbers

Image Properties Introduction Outline Image The smallest single component of a digital image is called a Processing Edges Pixel (Pixel = Picture Element) . Feature Detection SIFT Implementation of Code Neural Networks K-Means Color Cluster

Image Properties Introduction Outline Image The smallest single component of a digital image is called a Processing Edges Pixel (Pixel = Picture Element) . Feature Detection SIFT The resolution of an image is the number of pixels. Implementation of Code Neural Networks K-Means Color Cluster

Image Properties Introduction Outline Image The smallest single component of a digital image is called a Processing Edges Pixel (Pixel = Picture Element) . Feature Detection SIFT The resolution of an image is the number of pixels. Implementation of Code Neural e.g. A 12MP camera has a nominal 12,000,000 pixels. Networks K-Means Color Cluster

Image Properties Introduction Outline Image The smallest single component of a digital image is called a Processing Edges Pixel (Pixel = Picture Element) . Feature Detection SIFT The resolution of an image is the number of pixels. Implementation of Code Neural e.g. A 12MP camera has a nominal 12,000,000 pixels. Networks K-Means Color Cluster Image processing becomes more expensive as the resolution of the image increases.

What about images? Introduction Outline Image Processing Edges Feature Detection SIFT Implementation of Code Neural Networks K-Means Color Cluster

What about images? Introduction Outline Image Processing Edges Feature Detection SIFT Implementation of Code Neural Networks K-Means Color Cluster

Differentiation of Images Introduction Outline Given the following function, we can differentiate this function Image Processing in three different ways, Edges Feature Detection D ← f ( x ) = f ( x ) − f ( x − ∆ x ) SIFT ∆ x Implementation of Code Neural Networks K-Means Color Cluster

Differentiation of Images Introduction Outline Given the following function, we can differentiate this function Image Processing in three different ways, Edges Feature Detection D ← f ( x ) = f ( x ) − f ( x − ∆ x ) SIFT ∆ x Implementation of Code D → f ( x ) = f ( x + ∆ x ) − f ( x ) Neural Networks ∆ x K-Means Color Cluster

Differentiation of Images Introduction Outline Given the following function, we can differentiate this function Image Processing in three different ways, Edges Feature Detection D ← f ( x ) = f ( x ) − f ( x − ∆ x ) SIFT ∆ x Implementation of Code D → f ( x ) = f ( x + ∆ x ) − f ( x ) Neural Networks ∆ x K-Means D C f ( x ) = f ( x + ∆ x ) − f ( x − ∆ x ) Color Cluster ∆ x

Example of a Mask 5 Introduction Outline Image Processing Edges 4 Feature Detection SIFT Implementation 3 of Code Neural Networks 2 K-Means Color Cluster 1 0 1 2 3 4 5 6

Example of a Mask 5 Introduction Outline Image Processing 4 Edges Feature Detection SIFT 3 Implementation of Code Neural Networks 2 K-Means Color Cluster 1 0 1 2 3 4 5 6

Example of a Mask 5 Introduction Outline Image Processing 4 Edges Feature Detection SIFT 3 Implementation of Code Neural Networks 2 K-Means Color Cluster 1 0 1 2 3 4 5 6 Let ∆ x = 1, compute Df ← (3) , Df → (3) , Df C (3)

Sobel Edge Detection Introduction Outline Given an Image I ( x , y ), we want to extract where the changes Image Processing occur on the image. So what we do is convolve G ( x , y ) with Edges Feature Detection an operator, ∇ G ( x , y ), SIFT Implementation     − 1 0 1 − 1 − 2 − 1 of Code G x ( x , y ) = − 2 0 2 G y ( x , y ) = 0 0 0 Neural     Networks − 1 0 1 1 2 1 K-Means Color Cluster ( G x ( x , y )) 2 + ( G y ( x , y )) 2 � |∇ G ( x , y ) | =

contd...  1 1 2 2 2   0 0 0 0 0  Introduction 1 1 2 2 2 0 4 4 0 0 Outline     Image G x ( x , y )     Processing 1 1 2 2 2 − − − − → 0 4 4 0 0     Edges     Feature 1 1 2 2 2 0 4 4 0 0 Detection     1 1 2 2 2 0 0 0 0 0 SIFT Implementation of Code Neural     1 1 2 2 2 0 0 0 0 0 Networks 1 1 2 2 2 0 0 0 0 0 K-Means     Color Cluster G y ( x , y )     1 1 2 2 2 − − − − → 0 0 0 0 0         1 1 2 2 2 0 0 0 0 0     1 1 2 2 2 0 0 0 0 0

Feature Detection A feature is defined to be an “interesting” part of an image Introduction (edges, intensity changes, curves, lines, corners). There are Outline Image many different interesting feature detection methods: Processing Edges Feature Detection SIFT Implementation of Code Neural Networks K-Means Color Cluster

Feature Detection A feature is defined to be an “interesting” part of an image Introduction (edges, intensity changes, curves, lines, corners). There are Outline Image many different interesting feature detection methods: Processing Edges Feature Detection Edge Detection: Sobel, Canny, Canny-Deriche, Differential, SIFT Prewitt, Roberts cross Implementation of Code Neural Networks K-Means Color Cluster

Feature Detection A feature is defined to be an “interesting” part of an image Introduction (edges, intensity changes, curves, lines, corners). There are Outline Image many different interesting feature detection methods: Processing Edges Feature Detection Edge Detection: Sobel, Canny, Canny-Deriche, Differential, SIFT Prewitt, Roberts cross Implementation of Code Neural Blob Detection: Difference of Gaussian, Laplacian of Gaussian, Networks Determinant of Hessian K-Means Color Cluster

Feature Detection A feature is defined to be an “interesting” part of an image Introduction (edges, intensity changes, curves, lines, corners). There are Outline Image many different interesting feature detection methods: Processing Edges Feature Detection Edge Detection: Sobel, Canny, Canny-Deriche, Differential, SIFT Prewitt, Roberts cross Implementation of Code Neural Blob Detection: Difference of Gaussian, Laplacian of Gaussian, Networks Determinant of Hessian K-Means Color Cluster Line Detection: Hough Transform

Feature Detection A feature is defined to be an “interesting” part of an image Introduction (edges, intensity changes, curves, lines, corners). There are Outline Image many different interesting feature detection methods: Processing Edges Feature Detection Edge Detection: Sobel, Canny, Canny-Deriche, Differential, SIFT Prewitt, Roberts cross Implementation of Code Neural Blob Detection: Difference of Gaussian, Laplacian of Gaussian, Networks Determinant of Hessian K-Means Color Cluster Line Detection: Hough Transform Corner Detection: Harris

Formulation of the Problem Introduction Outline Image Scaling Processing Edges Feature Detection SIFT − → Implementation of Code Neural Networks K-Means Color Cluster

Formulation of the Problem Introduction Outline Image Scaling Processing Edges Feature Detection SIFT − → Implementation of Code Rotation Neural Networks K-Means Color Cluster − →