filtering images in the spatial domain chapter 3b g&wfiltering images in the spatial domain...
TRANSCRIPT
![Page 1: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/1.jpg)
1
Filtering Images in the Spatial Domain
Chapter 3b G&W Ross Whitaker
(modified by Guido Gerig) School of Computing
University of Utah
![Page 2: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/2.jpg)
2
Overview
• Correlation and convolution • Linear filtering
– Smoothing, kernels, models – Detection – Derivatives
• Nonlinear filtering – Median filtering – Bilateral filtering – Neighborhood statistics and nonlocal filtering
![Page 3: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/3.jpg)
3
0.0*95 + 0.1*103 + 0.0*150 + 0.1*36 + 0.6*150 + 0.1*104 + 0.0*47 + 0.1*205 + 0.0*77 = 134.8
. . . . . . . . . . . . . . . .
100 130 104 99 … 87 95 103 150 … 50 36 150 104 … 20 47 205 77 … . . . . . . . .
Filter
0.0 0.1 0.0 0.1 0.6 0.1 0.0 0.1 0.0
Input image Output image
0.0*87 + 0.1*95 + 0.0*103 + 0.1*50 + 0.6*36 + 0.1*150 + 0.0*20 + 0.1*47 + 0.0*205 = 55.8
Cross Correlation • Operation on image neighborhood and small …
– “mask”, “filter”, “stencil”, “kernel” • Linear operations within a moving window
55.8 134.8
![Page 4: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/4.jpg)
4
Cross Correlation
• 1D
• 2D
![Page 5: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/5.jpg)
5
Correlation: Technical Details
• How to filter boundary?
? ? ? ? ?
?
?
?
?
![Page 6: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/6.jpg)
6
Correlation: Technical Details • Boundary conditions
– Boundary not filtered (keep it 0) – Pad image with amount (a,b)
• Constant value or repeat edge values – Cyclical boundary conditions
• Wrap or mirroring
![Page 7: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/7.jpg)
7
Correlation: Technical Details
• Boundaries – Can also modify kernel – no longer
correlation • For analysis
– Image domains infinite – Data compact (goes to zero far away from
origin)
![Page 8: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/8.jpg)
8
Correlation: Properties
• Shift invariant
![Page 9: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/9.jpg)
9
Correlation: Properties
• Shift invariant
• Linear
Compact notation
![Page 10: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/10.jpg)
10
Filters: Considerations
• Normalize – Sums to one – Sums to zero (some cases, see later)
• Symmetry – Left, right, up, down – Rotational
• Special case: auto correlation
![Page 11: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/11.jpg)
11
0 0 0 0 1 0 0 0 0
1 1 1 1 1 1 1 1 1
1/9 *
Examples 1
![Page 12: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/12.jpg)
12
1 1 1 1 1 1 1 1 1
1/9 *
Examples 1
![Page 13: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/13.jpg)
13
1 1 1 1 1 1 1 1 1
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1/9 *
1/25 *
Examples 2
![Page 14: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/14.jpg)
14
Smoothing and Noise Noisy image 5x5 box filter
![Page 15: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/15.jpg)
15
Noise Analysis • Consider an a simple image I() with additive,
uncorrelated, zero-mean noise of variance s • What is the expected rms error of the
corrupted image? • If we process the image with a box filter of
size 2a+1 what is the expected error of the filtered image?
![Page 16: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/16.jpg)
17
Other Filters
• Disk – Circularly symmetric, jagged in discrete
case • Gaussians
– Circularly symmetric, smooth for large enough stdev
– Must normalize in order to sum to one • Derivatives – discrete/finite differences
– Operators
![Page 17: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/17.jpg)
Gaussian Kernel
18
![Page 18: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/18.jpg)
Gaussian Kernel
19
Normalization to 1.0
![Page 19: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/19.jpg)
20
• Discrete
• Continuous
• Same as cross correlation with kernel transposed around each axis
• The two operations (correlation and convolution) are the same if the kernel is symmetric about axes
Convolution
reflection of w
Java demo: http://www.jhu.edu/signals/convolve/ http://www.jhu.edu/signals/discreteconv2/index.html
![Page 20: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/20.jpg)
21
Convolution: Properties
• Shift invariant, linear • Commutative
• Associative
• Others (discussed later): – Derivatives, convolution theorem, spectrum…
![Page 21: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/21.jpg)
22
Computing Convolution
• Compute time – MxM mask – NxN image
O(M2N2) “for” loops are nested 4 deep
![Page 22: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/22.jpg)
23
Computing Convolution
• Compute time – MxM mask – NxN image
• Special case: separable
O(M2N2) “for” loops are nested 4 deep
O(M2N2) O(MN2)
Two 1D kernels
= *
![Page 23: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/23.jpg)
24
Separable Kernels
• Examples – Box/rectangle – Bilinear interpolation – Combinations of
partial derivatives • d2f/dxdy
– Gaussian • Only filter that is both
circularly symmetric and separable
• Counter examples – Disk – Cone – Pyramid
![Page 24: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/24.jpg)
25
Separability
![Page 25: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/25.jpg)
Box versus Gaussian
26
![Page 26: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/26.jpg)
Digital Images: Boundaries are “Lines” or “Discontinuities”
Example: Characterization of discontinuities? 27 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 27: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/27.jpg)
28 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 28: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/28.jpg)
29 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 29: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/29.jpg)
30
Derivatives: Finite Differences
![Page 30: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/30.jpg)
31
0 -1 0 0 0 0 0 1 0
0 0 0 -1 0 1 0 0 0
Derivative Example
![Page 31: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/31.jpg)
32 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 32: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/32.jpg)
33 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 33: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/33.jpg)
34 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 34: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/34.jpg)
35
Pattern Matching
![Page 35: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/35.jpg)
36
Pattern Matching/Detection
• The optimal (highest) response from a filter is the autocorrelation evaluated at position zero
• A filter responds best when it matches a pattern that looks itself
• Strategy – Detect objects in images by correlation with
“matched” filter
![Page 36: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/36.jpg)
37
Matched Filter Example
Trick: make sure kernel sums to zero
![Page 37: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/37.jpg)
38
Matched Filter Example: Correlation of template with image
![Page 38: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/38.jpg)
39
Matched Filter Example: Thresholding of correlation results
![Page 39: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/39.jpg)
40
Matched Filter Example: High correlation → template found
![Page 40: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/40.jpg)
44
Nonlinear Methods For Filtering
• Median filtering • Bilateral filtering • Neighborhood statistics and nonlocal
filtering
![Page 41: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/41.jpg)
45
Median Filtering
• For each neighborhood in image – Sliding window – Usually odd size (symmetric) 5x5, 7x7,…
• Sort the greyscale values • Set the center pixel to the median • Important: use “Jacobi” updates
– Separate input and output buffers – All statistics on the original image old new
![Page 42: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/42.jpg)
46
Median vs Gaussian
Original + Gaussian
Noise
3x3 Box 3x3 Median
![Page 43: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/43.jpg)
47
Median Filter
• Issues – Boundaries
• Compute on pixels that fall within window – Computational efficiency
• What is the best algorithm?
• Properties – Removes outliers (replacement noise – salt and
pepper) – Window size controls size of structures – Preserves straight edges, but rounds corners and
features
![Page 44: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/44.jpg)
48
Replacement Noise • Also: “shot noise”, “salt&pepper” • Replace certain % of pixels with samples from pdf • Best filtering strategy: filter to avoid outliers
![Page 45: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/45.jpg)
49
Smoothing of S&P Noise • It’s not zero mean (locally) • Averaging produces local biases
![Page 46: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/46.jpg)
50
Smoothing of S&P Noise • It’s not zero mean (locally) • Averaging produces local biases
![Page 47: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/47.jpg)
51
Median Filtering
Median 5x5 Median 3x3
![Page 48: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/48.jpg)
52
Median Filtering
Median 5x5 Median 3x3
![Page 49: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/49.jpg)
53
Median Filtering • Iterate
Median 3x3 2x Median 3x3
![Page 50: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/50.jpg)
54
Median Filtering
• Image model: piecewise constant (flat)
Ordering
Output
Ordering
Output
![Page 51: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/51.jpg)
55
Order Statistics • Median is special case of order-statistics filters • Instead of weights based on neighborhoods, weights
are based on ordering of data Neighborhood Ordering
Filter
Neighborhood average (box) Median filter
Trimmed average (outlier removal)
![Page 52: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/52.jpg)
56 Source: http://web.media.mit.edu/~maov/classes/vision09/lect/09_Image_Filtering_Edge_Detection_09.pdf
![Page 53: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/53.jpg)
57
Piecewise Flat Image Models
• Image piecewise flat -> average only within similar regions
• Problem: don’t know region boundaries
![Page 54: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/54.jpg)
58
Piecewise-Flat Image Models
• Assign probabilities to other pixels in the image belonging to the same region
• Two considerations – Distance: far away pixels are less likely to
be same region – Intensity: pixels with different intensities
are less likely to be same region
![Page 55: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/55.jpg)
59
Piecewise-Flat Images and Pixel Averaging
Distance (kernel/pdf) Distance (pdf)
Prob pixel belongs to same region as i
Prob pixel belongs to same region as i
intensity position
![Page 56: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/56.jpg)
60
• Neighborhood – sliding window • Weight contribution of neighbors according to:
• G is a Gaussian (or lowpass), as is H, N is neighborhood, – Often use G(rij) where rij is distance between pixels – Update must be normalized for the samples used in this
(particular) summation • Spatial Gaussian with extra weighting for intensity
– Weighted average in neighborhood with downgrading of intensity outliers
Bilateral Filter
Tomasi, Manduchi: http://en.wikipedia.org/wiki/Bilateral_filter http://homepages.inf.ed.ac.uk/rbf/CVonline/LOCAL_COPIES/MANDUCHI1/Bilateral_Filtering.html
normalization: all weights add up to 1
![Page 57: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/57.jpg)
Bilateral Filter
61
When the bilateral filter is centered, say, on a pixel on the bright side of the boundary, the similarity function s assumes values close to one for pixels on the same side, and values close to zero for pixels on the dark side. The similarity function is shown in figure 1(b) for a 23x23 filter support centered two pixels to the right of the step in figure 1(a).
Replaces the pixel value at x with an average of similar and nearby pixel values.
![Page 58: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/58.jpg)
62
Bilateral Filtering
Bilateral Gaussian Blurring
Replaces the pixel value at x with an average of similar and nearby pixel values.
![Page 59: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/59.jpg)
63
Bilateral Filtering
Bilateral Gaussian Blurring
![Page 60: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/60.jpg)
64
Nonlocal Averaging
• Recent algorithm – NL-means, Baudes et al., 2005 – UINTA, Awate & Whitaker, 2005
• Different model – No need for piecewise-flat – Images consist of some set of pixels with similar
neighborhoods → average several of those • Scattered around
– General area of a pixel – All around
• Idea – Average sets of pixels with similar neighborhoods
![Page 61: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/61.jpg)
UINTA: Unsupervised Information-Theoretic Adaptive Filtering : Excellent Introduction and
Additional Readings (Suyash P. Awate)
65
http://www.cs.utah.edu/~suyash/pubs/uinta/
Suyash P. Awate, Ross T. Whitaker Unsupervised, Information-Theoretic, Adaptive Image Filtering with Applications to Image Restoration IEEE Trans. Pattern Analysis & Machine Intelligence (TPAMI) 2006, Vol. 28, Num. 3, pp. 364-376
![Page 62: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/62.jpg)
66
Nonlocal Averaging
• Strategy: – Average pixels to alleviate noise – Combine pixels with similar neighborhoods
• Formulation – ni,j – vector of pixels values, indexed by j,
from neighborhood around pixel i
= – vector
![Page 63: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/63.jpg)
67
Nonlocal Averaging Formulation
• Distance between neighborhoods
• Kernel weights based on distances
• Pixel values of k neighborhoods: fk
22
2,
)( ,,s
kid
edKw kiki
-==
![Page 64: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/64.jpg)
68
Averaging Pixels Based on Weights
• For each pixel, i, choose a set of pixel locations k: – k = 1, …., M – Average them together based on
neighborhood weights (prop. to intensity pattern difference)
k
M
kkiwi fwg M
k kiå
=å¬
= 1,
1
1 ,
![Page 65: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/65.jpg)
69
Nonlocal Averaging
![Page 66: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/66.jpg)
70
Some Details
• Window sizes: good range is 5x5->11x11
• How to choose samples: – Random samples from around the image
• UINTA, Awate&Whitaker – Block around pixel (bigger than window,
e.g. 51x51) • NL-means
• Iterate – UNITA: smaller updates and iterate
![Page 67: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/67.jpg)
71
NL-Means Algorithm
• For each pixel, p – Loop over set of pixels nearby – Compare the neighorhoods of those pixels
to the neighborhood of p and construct a set of weights
– Replace the value of p with a weighted combination of values of other pixels
• Repeat… but 1 iteration is pretty good
![Page 68: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/68.jpg)
72
Results
Noisy image (range 0.0-1.0) Bilateral filter (3.0, 0.1)
![Page 69: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/69.jpg)
73
Results
Bilateral filter (3.0, 0.1) NL means (7, 31, 1.0)
![Page 70: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/70.jpg)
74
Results
Bilateral filter (3.0, 0.1) NL means (7, 31, 1.0)
![Page 71: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/71.jpg)
75
Less Noisy Example
![Page 72: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/72.jpg)
76
Less Noisy Example
![Page 73: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/73.jpg)
77
Results
Original Noisy Filtered
![Page 74: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/74.jpg)
78
Checkerboard With Noise
Original Noisy Filtered
![Page 75: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/75.jpg)
79
Quality of Denoising
• s, joint entropy, and RMS- error vs. number of iterations
![Page 76: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/76.jpg)
80
MRI Head
![Page 77: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/77.jpg)
81
MRI Head
![Page 78: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/78.jpg)
82
Fingerprint
![Page 79: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/79.jpg)
83
Fingerprint
![Page 80: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/80.jpg)
84
Results
Original Noisy Filtered
![Page 81: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/81.jpg)
85
Results
Original Noisy Filtered
![Page 82: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/82.jpg)
86
Results
Original Noisy Filtered
![Page 83: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/83.jpg)
87
Fractal
Original Noisy Filtered
![Page 84: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/84.jpg)
88
Piecewise Constant • Several 10s of Iterations • Tends to obliterate rare events (e.g. corners)
![Page 85: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/85.jpg)
89
Texture, Structure
![Page 86: Filtering Images in the Spatial Domain Chapter 3b G&WFiltering Images in the Spatial Domain Chapter 3b G&W Ross Whitaker (modified by Guido Gerig) School of Computing University of](https://reader033.vdocument.in/reader033/viewer/2022060715/607b25f9fd357f2ab3208b1f/html5/thumbnails/86.jpg)
90