Edge-based Segmentation Transform Hough Edge Tracking Linking - - PowerPoint PPT Presentation

▶
edge based segmentation
SMART_READER_LITE
LIVE PREVIEW

Edge-based Segmentation Transform Hough Edge Tracking Linking - - PowerPoint PPT Presentation

Sep 16, 2023 360 likes •692 views

Edge-based Marr-Hildreth http://vision.ouc.edu.cn/~zhenghaiyong CVBIOUC ZhaoHaiwei DaiJialun WangRuchen Edge-based Segmentation Transform Hough Edge Tracking Linking Edge Detection Canny Edge Techniques

slide-1
SLIDE 1

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge-based Segmentation

基于边缘的图像分割 ZhaoHaiwei DaiJialun WangRuchen

CVBIOUC http://vision.ouc.edu.cn/~zhenghaiyong

April 28, 2015

slide-2
SLIDE 2

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Contents

1 Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

2 Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

3 Edge Linking

Edge Tracking Hough Transform

slide-3
SLIDE 3

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

What?

slide-4
SLIDE 4

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge & Boundary

  • riginal image

edge boundary

The edge of object may be not a boundary, the boundary may also be not edge.

slide-5
SLIDE 5

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Problems

Edge detection is a simple and fast technique used in segmentation methods. However, there also are some problems: Noise and background afgect the accurate of edge detection. Edges are the sign of lack of continuity, and ending.

slide-6
SLIDE 6

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Steps in Edge Detection

1 Filtering: Filtering to reduce noise results in a loss of

edge strength.

2 Enhancement: In order to facilitate the detection of

edges, it is essential to determine changes in intensity in the neighborhood of a point.

3 Detection: Find the zero crossing and peak value to

detect edge.

slide-7
SLIDE 7

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Problems

Edge detection is a simple and fast technique used in segmentation methods. However, there also are some problems: Noise and background afgect the accurate of edge detection. Edges are the sign of lack of continuity, and ending.

slide-8
SLIDE 8

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Step in Image Segmentation

slide-9
SLIDE 9

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Detection

Three edge types and defjnitions: Step edge Ramp edge Roof edge

slide-10
SLIDE 10

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Detection

slide-11
SLIDE 11

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Detection Techniques1

Roberts Edge Detection Sobel Edge Detection Prewitt Edge Detection . . . Marr-Hildreth Edge Detection (LoG) Canny Edge Detection

1Rafael C. Gonzalez et al., “Digital Image Processing”, Publishing

House of Electronics Industry, 2011.

slide-12
SLIDE 12

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Gradient

The gradient ▽f of a function is: ▽f = grad(f ) = [ gx gy ] = [

Bf Bx Bf By

] The magnitude of the gradient is: M(x, y) = mag(▽f ) = b g2

x + g2 y

The direction of the greatest rate of change is: α(x, y) = arctan [gx gy ]

slide-13
SLIDE 13

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Detection Techniques

Roberts Edge Detection gx = Bf Bx = z9 ´ z5 gy = Bf By = z8 ´ z6

slide-14
SLIDE 14

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Detection Techniques

Sobel Edge Detection gx = Bf Bx = (z7 + 2z8 + z9) ´ (z1 + 2z2 + z3) gy = Bf By = (z3 + 2z6 + z9) ´ (z1 + 2z4 + z7) Prewitt Edge Detection

slide-15
SLIDE 15

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Marr-Hildreth Edge Detection (LoG)2

This method combines Gaussian fjltering with the Laplacian for edge detection. Idea:

1 Gaussian fjler is used to

swipe away noise from the image

2 Laplace operator 3 Locate zero crossings

2Marr D, Hildreth E, “Theory of Edge Detection”, Proc. of the

Royal Society of London, 1980.

slide-16
SLIDE 16

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection3

Canny is the best edge detection detector. Idea:

1 Low error rate 2 Edge points should be well localized

3Joun F. Canny, “A Computational Approach To Edge Detection”,

PAMI, 1986.

slide-17
SLIDE 17

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection

Canny algorithm:

slide-18
SLIDE 18

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection

Gauss Filter:

1 Gaussian distribution

  • ne dimension:

G(x) = 1 ? 2πσ e´ x2

2σ2

two dimension: G(x, y) = 1 2πσ2 e´ x2+y2

2σ2

2 Gaussian kernel

slide-19
SLIDE 19

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection

Compute Gradient: Sx = [ ´1 1 ´1 1 ] , Sy = [ 1 1 ´1 ´1 ]

slide-20
SLIDE 20

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection

Non-maxima Suppression (NMS): Edges generated using gradient typically contain wide ridges around local maxima. Use non-maxima suppression to thin those ridges to fjnd thin edges corresponding to local maxima.

slide-21
SLIDE 21

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection

slide-22
SLIDE 22

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Canny Edge Detection

Double thresholding: Problem: The received image may still contain false edge points.

slide-23
SLIDE 23

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Linking

Edge Tracking Hough Transform Curve Fitting Dynamic Programming . . .

slide-24
SLIDE 24

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Edge Tracking

Idea: Each point is linked to the adjacent if magnitude and direction of the gradient are similar. e(x, y) is magnitude of the gradient, θ(x, y) is the direction of the gradient, if two each points meet the following conditions： $ & % | e(xi, yi) ´ e(xj, yj) |ď E | θ(xi, yi) ´ θ(xj, yj) |ď A | e(xi, yi) |, | e(xj, yj) |ą E

slide-25
SLIDE 25

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Hough Transform

The Hough Transform can be used to detect lines, circles

  • r other parametric curves.

Idea:

slide-26
SLIDE 26

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Hough Transform

y = a

1x + b 1

" b = ´xia + yi b = ´xja + yj

slide-27
SLIDE 27

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Hough Transform

" ρ

1 = xicosθ 1 + yisinθ 1

ρ

1 = xjcosθ 1 + yjsinθ 1

ù ñ " xicosθ + yisinθ = ρ xjcosθ + yjsinθ = ρ

slide-28
SLIDE 28

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Hough Transform

Hough algorithm:

1 Quantize the parameter space (ρ, θ). This quantized

space is referred to as the accumulator cells.

2 Count the number of times a line intersects a given cell. 3 Lines can be found as peaks in this accumulator space.

ρ = xcosθ + ysinθ

slide-29
SLIDE 29

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Hough Transform

Circle: (x ´ a)2 + (y ´ b)2 = r

slide-30
SLIDE 30

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

(a ´ xi)2 + (b ´ yi)2 = r

slide-31
SLIDE 31

Edge-based Segmenta- tion Introduction

Edge & Boundary Problems Steps in Edge Detection Steps in Image Segmentation

Edge Detection

Edge Detection Techniques Marr-Hildreth Canny Edge Detection

Edge Linking

Edge Tracking Hough Transform

Thanks!