Low level feature extraction - chapter 4

52
Low-Level Feature Extraction By Alaa Mohammed Khattab 05/26/2022 1

Transcript of Low level feature extraction - chapter 4

PowerPoint Presentation

Low-Level Feature ExtractionByAlaa Mohammed Khattab12/27/20141

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/20142

Low Level Feature Extraction12/27/20143Edge DetectionMotion DetectionBasic features that can be extracted automatically from an image without any shape information (information about spatial relationships)

What is edge ?Edge defined as change or differences in intensity.12/27/20144

Types of edges:12/27/20145

Step (ideal) : Gray values change suddenly.Ramp : Grey values change slowly.

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/20146

Why ?Measure the size of the objects in an image.

Isolate particular objects from their background.

Recognize or classify objects.

12/27/20147

Edge detectionWe only consider changes in intensity in gray-scale images.

Edge detection are based on differentiation.

12/27/20148Image

Edge detection paradigm 12/27/20149

Determine changes in intensity in the neighborhood of a pointDetect points with strong edge content

Edge Magnitude & Direction12/27/201410

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/201411

Basic Operator12/27/201412

Roberts cross 196512/27/201413

Prewitt 196612/27/201414

Sobel 197012/27/201415

Canny 1986Canny edge detection is a four step process.Smoothing the image using Gaussian filter (reduce noise - false edges).

Apply Sobel or Perwitt operator.

Non-maximum suppression determines if the pixel is a better candidate for an edge than its neighbors (thinning).12/27/201416

Canny

Hysteresis thresholding finds where edges begin and end.12/27/201417

Canny12/27/201418

SobelCanny

CannyGood detection the algorithm should mark as many real edges in the image as possible.

Works fine under noisy condition.

Good localization edges marked should be as close as possible to the edge in the real image.

Minimal response a given edge in the image should only be marked once, and where possible, image noise should not create false edges.

12/27/201419

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/201420

Second-Order edge detectionZero-crossing of the second derivative of a function indicates the presence of an edge.12/27/201421

Laplacian12/27/201422

Zero Crossing12/27/201423

Marr-Hildreth 1980 (LoG)Smooth the image with a Gaussian filter.Enhance the edges using Laplacian operator.Find the zero-crossing.12/27/201424

CannyLOG

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/201425

Spacek 198612/27/201426

Petrou 1991Petrou questioned the validity of the step edge model for real images.

Any step-changes in the image will be smoothed to become a ramp.

Petrou operator uses templates that are 12 pixels wide at minimum, in order to preserve optimal properties.12/27/201427

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/201428

Comparison of First-Order Operators12/27/201429

Original ImageRobertsPrewittSobelCanny

Comparison of edge detection operators12/27/201430

CannySpacekPetrou

Comparison of edge detection operators12/27/201431

Ultrasound imageBasic operatorPrewittSobelMarr-HildrethCannySpacek

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/201432

Describing image motionMotion detection based on comparing of the current video frame with one from the previous frames.12/27/201433

xytime

Describing image motion12/27/201434

12/27/201435

Second imageFirst imageDifference image

Describing image motionThere are a lot of approaches to detect the motion in the image such as:Area-based approach.Differential approach.12/27/201436

Area-based approach12/27/201437

time ttime t+1

Area-based approach

Motion can be characterized as a collection of displacements in the image plane.

We try to find correlation between the pixel in the old and new frames.12/27/201438

Area-based approach12/27/201439

Area-based approach12/27/201440

Area-based approachAssume that:The brightness at the point in the new position should be the same as the brightness at the old position.

The neighboring points move with similar velocity.12/27/201441

Describing image motionThere are a lot of approaches to detect the motion in the image such as:Area-based approach.Differential approach.12/27/201442

Differential approach12/27/201443

Differential approach12/27/201444

Differential approach12/27/201445

Differential approach12/27/201446

Correlation vs. DifferentialArea-basedDifferential

Slow.high computation.

Flow is clear.

Not concerned with rotation.

Faster than area-based.

Uncertain flow.12/27/201447

Correlation vs. DifferentialArea-basedDifferential

12/27/201448

ContentIntroduction.Edge Detection.First-Order edge detection.( Basic , Roberts , Prewitt , Sobel , Canny )Second-Order edge detection.( Zero-crossing , Marr-Hildreth)Other edge detection operators.( Spacek , Petrou )Comparison of edge detection operators.Describing image motion.( Area-based , Differential approach )Implementation.12/27/201449

Implementation (MatLab)First-Order Edge detectionRoberts:edge ( I , roberts , thresh , options )Prewitt:edge ( I , prewitt , thresh , direction )Sobel:edge ( I , sobel , thresh , options , direction )Canny:edge ( I , canny , sigma )12/27/201450

Implementation (MatLab)Second-Order Edge detectionZero-Crossing:edge ( I , zerocross , h , thresh )

Marr-Hildreth:edge ( I , log , thresh , sigma )12/27/201451

12/27/201452