lecture 9.2 image feature extraction€¦ · feature extraction . the goal is to generate features...
TRANSCRIPT
![Page 1: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/1.jpg)
Lecture 9.2 Image Feature Extraction
Idar Dyrdal
![Page 2: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/2.jpg)
Classification system
2
Camera Feature Extractor
(Image Processing)
Classifier
![Page 3: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/3.jpg)
Image Analysis
Typical steps: • Pre-processing • Segmentation (object detection) • Feature extraction • Feature selection • Classifier training • Evaluation of classifier performance.
3
![Page 4: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/4.jpg)
Features for image analysis
Applications: • Remote sensing • Medical imaging • Character recognition • Robot Vision • …
Major goal of image feature extraction:
Given an image, or a region within an image, generate the features that will subsequently be fed to a classifier in order to classify the image in one of the possible classes. (Theodoridis & Koutroumbas: «Pattern Recognition», Elsevier 2006).
4
![Page 5: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/5.jpg)
Feature extraction
The goal is to generate features that exhibit high information-packing properties: • Extract the information from the raw data that is most relevant for
discrimination between the classes • Extract features with low within-class variability and high between
class variability • Discard redundant information. • The information in an image f[i,j] must be reduced to enable reliable
classification (generalization) • A 64x64 image à 4096-dimensional feature space!
5
![Page 6: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/6.jpg)
“Curse of dimensionality”
6
New data
Training data
Error rate
d
![Page 7: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/7.jpg)
Feature types (regional features)
• Colour features • Gray level features • Shape features • Histogram (texture) features
7
![Page 8: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/8.jpg)
Shape features - example
8
![Page 9: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/9.jpg)
Moments
9
Geometric moments (order p,q):
Central moments:
![Page 10: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/10.jpg)
Binary images
10
![Page 11: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/11.jpg)
Moments of inertia
11
![Page 12: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/12.jpg)
Closest fitting ellipse
12
Orientation:
Eccentrisity:
x
y
![Page 13: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/13.jpg)
Major an minor axes
13
![Page 14: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/14.jpg)
Histogram (texture) features
• First order statistics (information related to the gray level distribution) • Second order statistics (information related to spatial/relative distribution of gray level), i.e.
second order histogram, co-occurrence matrix Histogram: Moments from gray level histogram: Entropy:
14
![Page 15: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/15.jpg)
Histogram (texture) features
Central moments:
Features:
15
![Page 16: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/16.jpg)
Feature selection
• A number of feature candidates may have been generated
• Using all candidates will easily lead to over traing (unreliable classification of new data)
• Dimmensionality reduction is required, i.e. feature selection!
• Exhaustive search impossible! • Trial and error (select feature combination,
train classifier, estimate error rate). • Suboptimal search • «Branch and Bound» search • Linear or non-linear mappings to lower
dimensional feature space.
Scatter plot of features
![Page 17: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/17.jpg)
Dimensionality reduction – linear transformations
17
• Projection of multidimensional feature vectors to a lower-dimensional feature space
• Example: Fishers linear discriminant provides a projection from a d-dimensional space (d>1) to a one-dimensional space in such a way that the separation between classes are maximized.
![Page 18: Lecture 9.2 Image Feature Extraction€¦ · Feature extraction . The goal is to generate features that exhibit high information- packing properties: • Extract the information from](https://reader033.vdocument.in/reader033/viewer/2022051907/5ff9c30dd7955302dc67c8da/html5/thumbnails/18.jpg)
Summary
Image feature extraction: • Feature extraction • Feature selection
Read also: Szeliski 14.4
18