EDGE DETECTION - PowerPoint PPT Presentation

About This Presentation
Title:

EDGE DETECTION

Description:

Isotropic. Good Localization. Good Signal-To-Noise Characteristics ... PROPERTIES OF A GOOD EDGE DETECTOR. September 23, 1998. 6. ISOTROPIC EDGE DETECTION ... – PowerPoint PPT presentation

Number of Views:88
Avg rating:3.0/5.0
Slides: 21
Provided by: larry105
Learn more at: https://www.cs.jhu.edu
Category:

less

Transcript and Presenter's Notes

Title: EDGE DETECTION


1
EDGE DETECTION
2
IMAGE HEIGHT SURFACE
Graylevel
I(x,y)
Y
IMAGE PLANE
X
3
STEP EDGE
Gray_2
Gray_1
Sideview
4
FIRST ORDER EDGE DETECTION
Prewitt Templates
5
PROPERTIES OF A GOOD EDGE DETECTOR
  • Isotropic
  • Good Localization
  • Good Signal-To-Noise Characteristics
  • Accurate Determination of Edge Orientation

6
ISOTROPIC EDGE DETECTION
7
SIGNAL-TO-NOISE RATIO AND LOCALIZATION
8
PATTERNS OF VARIOUS FREQUENCIES
9
DERIVATIVES OF A STEP-EDGE
10
SECOND ORDER EDGE OPERATOR
MARR-HILDRETH OPERATOR
Combine this with Gaussian smoothing.
Detect edge at Zero-Crossing looking for
positive and negative peaks on either side.
11
GAUSSIAN CURVES IN 1-D AND 2-D
12
MEXICAN HAT FILTER
13
SECOND ORDER EDGE OPERATOR
CANNY OPERATOR
Take second derivative in direction of gradient
(Ix,Iy). Look for Zero-Crossing of second
derivative in this direction. Can also
threshold with respect to first order edge
strength. Direction of first order gradient
(Ix,Iy) gives edge orientation. Proved to have
optimal localization properties.
14
CANNY EDGE OPERATOR
  • Only consider pixel points with first order
    edge-strength above thresh_1.
  • Look for zero-crossings of second derivative in
    direction of image gradient.
  • Nonmaximum Supression Only consider those
    zero-crossings that constitute a maximum of the
    image intensity gradient.
  • Edge Tracking Pick thresh_2ltthresh_1 and travel
    perpendicular to gradient selecting pixels with
    image intensity gradient gtthresh_2.

15
IMAGE COMPRESSION
  • Compression Ratio
  • (Uncompressed File Size)/(Compressed File Size)
  • Bits per Pixel
  • (Number of bits)/(Number of pixels)

16
OBJECTIVE METRICS FOR IMAGE COMPRESSION
17
RUN LENGTH ENCODING(BINARY)
8, 4, 5, 2, 6, 5, 3,3,2,1,3,1,2,2,1,4,1,1,10
18
RUN LENGTH ENCODING(GRAYLEVEL)
1
0
1
1
1
0
0
1
BITPLANES
19
LZW COMPRESSION
  • Uses a code table to encode reoccurring strings
    of pixel values.
  • Example Consider a grey-level image with 8 bits
    per pixel. Make a code table of 12-bit words
    (therefore 4096 entries) where first 256 entries
    are individual grey values and next 3840 entries
    encode strings of grey values.

20
QUAD-TREE REPRESENTATION
Write a Comment
User Comments (0)
About PowerShow.com