Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020....
Transcript of Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020....
![Page 1: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/1.jpg)
Digital Image ProcessingCOSC 6380/4393
Lecture – 6
Sept 10th, 2020
Pranav MantiniSlides from Dr. Shishir K Shah and Frank (Qingzhong) Liu
![Page 2: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/2.jpg)
Resampling
• Once the image is acquired.
• How to
– Enlarge an image
– Shrink an image
– Zoom in
• Zooming Example:
– Initial image size = 500 X 500
– Required image size (= X 1.5) = 750 X 750
![Page 3: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/3.jpg)
Weeks 1 & 2 3
Image Interpolation
• Interpolation — Process of using known data to estimate unknown values
e.g., zooming, shrinking, rotating, and geometric correction
• Interpolation (sometimes called resampling) — an imaging method to increase (or decrease) the number of pixels in a digital image.
Some digital cameras use interpolation to produce a larger image than the
sensor captured or to create digital zoom
http://www.dpreview.com/learn/?/key=interpolation
![Page 4: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/4.jpg)
Interpolation: Nearest Neighbor
5 X 8 Image3 X 3 Image
Fill in values preserving spatial relationship
![Page 5: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/5.jpg)
Linear Interpolation
Solve for 𝐼𝐼2 − 𝐼1𝐼 − 𝐼1
=𝑥2 − 𝑥1𝑥 − 𝑥1
(𝐼2−𝐼1)𝑥 − 𝑥1𝑥2 − 𝑥1
= 𝐼 − 𝐼1
𝐼 = 𝐼1 + (𝐼2−𝐼1)𝑥 − 𝑥1𝑥2 − 𝑥1
𝑥1 𝑥2
𝐼1 𝐼2
𝑥
𝐼
![Page 6: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/6.jpg)
Bi-Linear Interpolation(2D)Q11 = (x1, y1),Q12 = (x1, y2),Q21 = (x2, y1),and Q22 = (x2, y2)
Find the value at 𝑃
https://en.wikipedia.org/wiki/Bilinear_interpolation
![Page 7: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/7.jpg)
Review: DIGITAL IMAGE REPRESENTATION
• Once an image is digitized (A/D) and stored it is an array of voltage or magnetic potentials
• Not easy to work with from an algorithmic point of view
• The representation that is easiest to work with from an algorithmic perspective is that of a matrix of integers
Matrix Image Representation• Denote a (square) image matrix I = [I(i, j); 0 < i, j < N-1]
where
• (i, j) = (row, column)
• I(i, j) = image value at coordinate or pixel (i, j)
![Page 8: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/8.jpg)
Review: DIGITAL IMAGE REPRESENTATION
(contd.)
• Example - Matrix notation
• Example - Pixel notation - an N x N image
What’s the minimum number
of bits/pixel allocated?
![Page 9: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/9.jpg)
Review: DIGITAL IMAGE REPRESENTATION
(contd.)• Example - Binary Image
(2-valued, usually
BLACK and WHITE)
• Another way of depicting the
image:
![Page 10: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/10.jpg)
10
BINARY IMAGES
• Since binary = bi-valued, the (logical) values ’0’ or ’1’ usually indicate the
absence or presence of an image property in an associated gray-level
image:
– Points of high or low intensity (brightness)
– Points where an object is present or absent
– More abstract properties, such as smooth vs. nonsmooth, etc.
• Convention - We will make the associations
• ’1’ = BLACK
• ’0’ = WHITE
![Page 11: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/11.jpg)
11
BINARY IMAGE GENERATION
• Tablet-Based Input:
• Binary images can derive from simple sensors with binary output
• Simplest example: tablet, resistive pad, or light pen
• All pixels initially assigned value ’0’:
I = [I(i, j)], I(i, j) = ’0’ for all (i, j) = (row column)
• When pressure or light is applied at (i0 , j0 ), the image is assigned the value ’1’: I(i0 , j0 ) = ’1’
• This continues until the user completes the drawing
![Page 12: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/12.jpg)
12
BINARY IMAGE
• Usually a binary image is obtained from a gray-level image
• Advantages:– B-fold reduction in required storage
– Simple abstraction of information
– Fast processing - logical operators
– Can be further compressed
![Page 13: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/13.jpg)
Grey Level Binary Image
8X8 image Black box on white background
![Page 14: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/14.jpg)
Grey Level Binary Image
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Grey scale Pixels values
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Binary image
![Page 15: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/15.jpg)
15
GRAY-LEVEL THRESHOLDING
Simple Thresholding
• The simplest of image processing operations
• An extreme form of gray-level quantization
• Define an integer threshold T (in the gray-scale range)
• Compare each pixel intensity to T
![Page 16: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/16.jpg)
16
THRESHOLDING
• Suppose gray-level image I has K gray-levels: 0, 1, 2, ...., K-1
• Select threshold T { 0, 1, 2, ...., K-1}
• Compare every gray-level in I to T
• Define a new binary image J as follows:
• J(i, j) = ’0’ if I(i, j) <= T
• J(i, j) = '1' if I(i, j) > T
• A new binary image J is created from a gray-level image I
![Page 17: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/17.jpg)
Grey Level Binary Image
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Grey scale Pixels values
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Binary image
𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑(T)
![Page 18: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/18.jpg)
Grey Level Binary Image
Grey scale Pixels values
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 1 1 1 1 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Binary image
𝑇ℎ𝑟𝑒𝑠ℎ𝑜𝑙𝑑(T)
What is good value of T?
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 255 255 255 255 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
![Page 19: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/19.jpg)
Grey Level Binary Image
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
8X8 image grey box on black background
What is good value of T?
![Page 20: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/20.jpg)
Grey Level Binary Image
8X8 image grey box on black background
8X8 image white box on dark white background
What is good value of T? What is good value of T?
240 240 240 240 240 240 240 240
240 240 240 240 240 240 240 240
240 240 255 255 255 255 240 240
240 240 255 255 255 255 240 240
240 240 255 255 255 255 240 240
240 240 255 255 255 255 240 240
240 240 240 240 240 240 240 240
240 240 240 240 240 240 240 240
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
![Page 21: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/21.jpg)
21
THRESHOLD SELECTION
• The quality of the binary image J obtained by thresholding I
depends very heavily on the threshold T
• Indeed it is instructive to observe the result of thresholding an
image at many different levels in sequence
• Different thresholds can produce different valuable abstractions
of the image
• Some images do not produce any interesting results when
thresholded by any T
![Page 22: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/22.jpg)
Grey Level Binary Image
8X8 image grey box on black background
8X8 image white box on dark white background
What is good value of T? What is good value of T?
240 240 240 240 240 240 240 240
240 240 240 240 240 240 240 240
240 240 255 255 255 255 240 240
240 240 255 255 255 255 240 240
240 240 255 255 255 255 240 240
240 240 255 255 255 255 240 240
240 240 240 240 240 240 240 240
240 240 240 240 240 240 240 240
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 10 10 10 10 0 0
0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
![Page 23: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/23.jpg)
23
THRESHOLD SELECTION
• The quality of the binary image J obtained by thresholding I
depends very heavily on the threshold T
• Indeed it is instructive to observe the result of thresholding an
image at many different levels in sequence
• Different thresholds can produce different valuable abstractions
of the image
• Some images do not produce any interesting results when
thresholded by any T
• So: How does one decide if thresholding is possible ?
• How does one decide on a threshold T ?
![Page 24: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/24.jpg)
Grey Level Binary Image
8X8 image light white box on white background
How do we determine 𝑇?
10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
8X8 image black box on grey background
![Page 25: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/25.jpg)
Determine modes
0
20
40
60
0 1 …
24
0
24
1
24
2 …
25
5
Pixel Count
0
20
40
60
0 1 2 …. 10 11 12 …
Pixel Count
10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
![Page 26: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/26.jpg)
Determine modes
𝑚𝑜𝑑𝑒1 = 0;𝑚𝑜𝑑𝑒2 = 10𝑇 = 𝑎𝑣𝑔 𝑚𝑜𝑑𝑒 = 𝟓
𝑚𝑜𝑑𝑒1 = 240;𝑚𝑜𝑑𝑒2 = 255𝑇 = 𝑎𝑣𝑔 𝑚𝑜𝑑𝑒 = 𝟐𝟒𝟕.5
10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 0 0 0 0 10 10
10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 240 240 240 240 255 255
255 255 255 255 255 255 255 255
255 255 255 255 255 255 255 255
0
20
40
60
0 1 …
24
0
24
1
24
2 …
25
5
Pixel Count
0
20
40
60
0 1 2 …. 10 11 12 …
Pixel Count
![Page 27: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/27.jpg)
27
GRAY-LEVEL IMAGE HISTOGRAM
• The histogram HI of image I is a plot or graph of the
frequency of occurrence of each gray level in I
• HI is a one-dimensional function with domain 0, ... , K-1
• HI(x) = n if I contains exactly n occurrences of gray level
x, for each x = 0, ... K-1
![Page 28: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/28.jpg)
Histogram Example
input imageintensity
pix
el
co
un
t
Black = 0
Gray = 190
White = 254
![Page 29: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/29.jpg)
29
HISTOGRAM APPEARANCE
• The appearance of a histogram suggests much about the image
![Page 30: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/30.jpg)
30
HISTOGRAM APPEARANCE
• The appearance of a histogram suggests much about the image
• These could be histograms of underexposed and overexposed images, respectively
predominantly
dark imagepredominantly
light image
![Page 31: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/31.jpg)
31
HISTOGRAM APPEARANCE
• This histogram may show better use of the gray-scale range
• Well-distributed histogram
![Page 32: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/32.jpg)
Histogram Example
input imageintensity
pix
el
co
un
t
Black = 0
Gray = 190
White = 254
![Page 33: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/33.jpg)
Histogram Example
Prof. Shishir Shah 33
input image
• Reality
intensity
pix
el
co
un
t
![Page 34: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/34.jpg)
34
BIMODAL HISTOGRAM
• Thresholding usually works best when there are dark objects on a light background
• Or when there are light objects on a dark background
• Images of this type tend to have histograms with multiple distinct peaks or modes in them
![Page 35: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/35.jpg)
35
BIMODAL HISTOGRAM
• If the peaks are well-separated, threshold selection can be easy
• Set the threshold T somewhere between the peaks
• It may be an interactive trial-and-error process
bimodal histogram
poorly separatedbimodal histogram
well separated peaks
![Page 36: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/36.jpg)
36
THRESHOLD SELECTION FROM
HISTOGRAM• Placing threshold T between
modes may yield acceptable
results
• Exactly where in between
can be difficult to determine
• An image histogram may
contain multiple modes.
Placing the threshold in
different places will produce
very different results
• Histogram may be "flat,"
making threshold selection
difficult
![Page 37: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/37.jpg)
Example: How to find T
• Microscopic image
• Grey level binary image
• Binary: 1-cell present, 0-cell absent
![Page 38: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/38.jpg)
Example: How to find T
• Determine peaks
• Choose 𝑇 between peaks(say average)
𝑝𝑒𝑎𝑘1
𝑻
𝑝𝑒𝑎𝑘2
![Page 39: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/39.jpg)
Example: How to find T
• Determine peaks
• Choose 𝑇 between peaks(say average)
𝑻
How to estimate/ approximate peaks?
𝑝𝑒𝑎𝑘1 𝑝𝑒𝑎𝑘2
![Page 40: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/40.jpg)
Recap: Probability
• Data: {5,5,6,6,6,7,7,7,7,7,7,8,8,8,9,9,10,10}
• 𝑋: random variable
• P: X [0,1] probability function
• 𝑃 𝑋 = 7 = ?
![Page 41: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/41.jpg)
Recap: Probability
• Data: {5,5,6,6,6,7,7,7,7,7,7,8,8,8,9,9,10,10}
• 𝑋: random variable
• P: probability function
• 𝑃(𝑋 = 7) = 0.33
![Page 42: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/42.jpg)
Recap: Histogram
• Data: {5,5,6,6,6,7,7,7,7,7,7,8,8,8,9,9,10,10}
• 𝑋: random variable
• P: probability function
• 𝑃(𝑋 = 7) = 0.33
• Histogram0
2
4
6
8
1 2 3 4 5 6 7 8 9 10 11 12
freq.
![Page 43: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/43.jpg)
Recap: Probability Distribution
• Data: {5,5,6,6,6,7,7,7,7,7,7,8,8,8,9,9,10,10}
• 𝑋: random variable
• P: probability function
• 𝑃(𝑋 = 7) = 0.33
• PNormalize(Histogram)
• What is a single value that represents this distribution
0
1/10
1/5
3/10
2/5
1 2 3 4 5 6 7 8 9 10 11 12
P
Probability distribution
![Page 44: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/44.jpg)
Expectation
• 𝐸(𝑋)Expected value of random variable X
• ~Average of all the expected values of random variable X
• 𝐸 𝑋 =?
![Page 45: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/45.jpg)
Expectation
• 𝐸(𝑋)Expected value of random variable X
• ~Average of all the expected values of random variable X
• 𝐸 𝑋 = σ𝑋𝑃(𝑋)
• Example: X outcome of rolling a dice
• Sample space: {1,2,3,4,5,6}
• P(X=1) = 1/6, P(X=2) = 1/6, …
• E[X] = ?
![Page 46: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/46.jpg)
Expectation
• 𝐸(𝑋)Expected value of random variable X
• ~Average of all the expected values of random variable X
• 𝐸 𝑋 = σ𝑋𝑃(𝑋)
• Example: X outcome of rolling a dice
• Sample space: {1,2,3,4,5,6}
• P(X=1) = 1/6, P(X=2) = 1/6, …
• E[X] = 1*1/6 + 2*1/6 +…. = 3.5
![Page 47: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/47.jpg)
Recap: Probability Distribution
• Data: {5,5,6,6,6,7,7,7,7,7,7,8,8,8,9,9,10,10}• 𝑋: random variable• P: probability function
• 𝑃(𝑋 = 7) = 0.33• PNormalize(Histogram)
• How to represent this data using a single value?• E(X) = 5 * 0.11 + 6*0.17 + 7*0.33 +… = 7.33
0
1/10
1/5
3/10
2/5
1 2 3 4 5 6 7 8 9 10 11 12
P
Probability distribution
![Page 48: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/48.jpg)
Example: How to find T
• Determine peaks
• Choose 𝑇 between peaks(say average)
𝑻
How to estimate/ approximate peaks?
𝑝𝑒𝑎𝑘1 𝑝𝑒𝑎𝑘2
![Page 49: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/49.jpg)
49
Algorithm
𝐼𝑛𝑖𝑡𝑖𝑎𝑙𝑖𝑧𝑒 𝑇 = 𝐾/2𝐷𝑜
𝐶𝑜𝑚𝑝𝑢𝑡𝑒 μ1 = 𝐸 𝑋 ∀𝑋 < 𝑇𝐶𝑜𝑚𝑝𝑢𝑡𝑒 μ2 = 𝐸 𝑋 ∀𝑋 ≥ 𝑇
𝑆𝑒𝑡 𝑇 =𝜇1 + 𝜇2
2𝑊ℎ𝑖𝑙𝑒 ∆𝜇1! = 0 &∆𝜇2! = 0
bimodal histogram
well separated peaks
AKA: Expectation Maximization (simple version)
![Page 50: Digital Image Processing · Digital Image Processing COSC 6380/4393 Lecture – 6 Jan 30 th, 2020. Pranav Mantini. Slides from Dr. Shishir K Shah and Frank (Qingzhong) Liu](https://reader036.fdocuments.in/reader036/viewer/2022071211/60233f898da7ee535f360d53/html5/thumbnails/50.jpg)
50
BIMODAL HISTOGRAM
bimodal histogram
poorly separatedbimodal histogram
well separated peaks