12
Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

Embed Size (px)

DESCRIPTION

Analog vs Digital  Analog Continuous value at continuous locations Used in differential-integral math  Digital Continuous/discrete value at discrete locations Used in linear algebra math I(x, y) at (x, y) I(m, n) at (m, n), m = [1, M], n = [1, N]

Citation preview

Page 1: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

Image Representation

Last update 2015. 1st MarchHeejune Ahn, SeoulTech

Page 2: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

1. Image Digital Image

Pixels, 2-Dimensional function of sensor value Origin: top-left, not bottom-left m (row index, vertical, i.e., y), n (col. Index,

horizontal, ie. x) 3-D image

voxels E.g. medical, 3D scanning

I(m, n) at (m, n), m = [1, M], n = [1, N]

I(0,n)

I(m,0) I(m,n)

Page 3: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

Analog vs Digital Analog

Continuous value at continuous locations Used in differential-integral math

Digital

Continuous/discrete value at discrete locations Used in linear algebra math

I(x, y) at (x, y)

I(m, n) at (m, n), m = [1, M], n = [1, N]

Page 4: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

2. Image values # of values/channels :1 to many Single value

gray scale Color map

Gray map : [0, max] to [black to white] False color map : [0. max] to “many colors”

Medical, astronomical application Better recognition (HVS: limited only 40 contrast

levels) E.g.) Jet-color map

3 channels (R,G,B), (H,S,V)

[0, max]

Color-map Color display (R, G, B)

Page 5: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

2. Resolution Accuracy of data Spatial resolution: # of pixels, e.g. MxN in 2D Temporal resolution: fps (frame/sec) Bit resolution (dynamic range): 1 bit, 8 bits,

24 bits, floating points.

Page 6: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

Real image resolution Representation vs real resolution How to determine image’s resolution Useful for real-time implementation &

performance

Page 7: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

3. Image file format Image file

File header + image values

Types file type, resolution, compression, etc

Page 8: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

Image data types Binary image: value = {0, 1}

0: black, background, 1: white, foreground Often mapped into [0, 255] E.g. Fax, resultant image

Gray-scale (intensity): [0,255] R,G,B true color: I[m,n, channel] channel=1,2,3 Floating point types

scientific & medical image, e.g. TIFF, medical DICOM

Page 9: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech
Page 10: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

4. Color spaces RGB

[min, max] to [0, 1] RGB to gray-scale

IGRAY(n,m) = IR(n,m) + IG(n,m) + IB(n,m)

IGRAY(n,m)=0.2989IR(n,m)+0.5870IG(n,m)+0.1140 IB(n,m) Gray to RGB is impossible (irreversible)

HSV perceived luminance physical power

Page 11: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

HSV (perceptual color space) H (hue) : dominant wavelength S (saturation) : the purity of color V (value) : brightness/luminance Less sensitive to lighting condition than RGB In Matlab : (0, 1)

Page 12: Image Representation Last update 2015. 1 st March Heejune Ahn, SeoulTech

5. Image in Matlab “im” prefix

image toolbox function starts with it. Read, write, & Query image files

iminfo, imwrite, imread Basic display

imshow: image data (value range [0,255]) imagesc: auto scaling/color map(any matrix input)

accessing pixel value imview

Converting types rgb2gray, rgb2hsv, hsv2rgb