Digital Image Processing Lecture 5 Neighborhood Processing Spatial
![Digital Image Processing Lecture 5: Neighborhood Processing: Spatial Filtering March 9, 2004 Prof. Charlene Digital Image Processing Lecture 5: Neighborhood Processing: Spatial Filtering March 9, 2004 Prof. Charlene](https://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-1.jpg)
Digital Image Processing Lecture 5: Neighborhood Processing: Spatial Filtering March 9, 2004 Prof. Charlene Tsai Digital Image Processing Lecture 5
![Spatial Filtering § Definition: a process that moves a subimage from point to point Spatial Filtering § Definition: a process that moves a subimage from point to point](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-2.jpg)
Spatial Filtering § Definition: a process that moves a subimage from point to point in an image, with the response at each image point predefined. § The subimage: filter, mask, kernel, template or window. § The values in a filter subimage are called coefficients, not pixels. § The process is also named convolution. § Convolution of 2 D function and is denoted or Digital Image Processing Lecture 5 2
![Illustration y origin W(-1, -1) W(-1, 0) W(-1, 1) mask W(0, -1) W(0, 0) Illustration y origin W(-1, -1) W(-1, 0) W(-1, 1) mask W(0, -1) W(0, 0)](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-3.jpg)
Illustration y origin W(-1, -1) W(-1, 0) W(-1, 1) mask W(0, -1) W(0, 0) W(0, 1) W(1, -1) W(1, 0) W(1, 1) f(x-1, y-1) f(x-1, y+1) f(x, y-1) f(x, y+1) Mask coefficients x f(x+1, y-1) f(x+1, y+1) Image section under mask Digital Image Processing Lecture 5 3
![Example 1: Averaging § One simple example is smoothing using a 3 x 3 Example 1: Averaging § One simple example is smoothing using a 3 x 3](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-4.jpg)
Example 1: Averaging § One simple example is smoothing using a 3 x 3 mask. f(x-1, y-1) f(x-1, y+1) f(x, y-1) f(x, y+1) f(x+1, y-1) f(x+1, y+1) Digital Image Processing Lecture 5 1/9 1/9 1/9 4
![Example 2: Delta/Impulse Funciton § An ideal impulse is defined using the Dirac distribution Example 2: Delta/Impulse Funciton § An ideal impulse is defined using the Dirac distribution](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-5.jpg)
Example 2: Delta/Impulse Funciton § An ideal impulse is defined using the Dirac distribution § To visualize in 1 D, picture a rectancular pulse from to with a height of. As , the width tends to 0 and the height tends to infinity as the total area remains constant at 1. 1 Digital Image Processing Lecture 5 5
![Example 2: Delta/Impulse Function (con’d) § Sifting property: It provides the value of f(x, Example 2: Delta/Impulse Function (con’d) § Sifting property: It provides the value of f(x,](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-6.jpg)
Example 2: Delta/Impulse Function (con’d) § Sifting property: It provides the value of f(x, y) at the point (a, b) § Delta function as the mask. § § When moving across an image, it “copies” the value of image intensity. Digital Image Processing Lecture 5 6
![General Formulation § 2 D Continuous space: § 2 D discrete space: Convolution kernel/mask General Formulation § 2 D Continuous space: § 2 D discrete space: Convolution kernel/mask](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-7.jpg)
General Formulation § 2 D Continuous space: § 2 D discrete space: Convolution kernel/mask § In digital image processing, we only deal with discrete space with a mask effective locally. § Convolution is commutative, associative and distributive. Digital Image Processing Lecture 5 7
![What happens at the borders? § The mask falls outside the edge. § Solutions? What happens at the borders? § The mask falls outside the edge. § Solutions?](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-8.jpg)
What happens at the borders? § The mask falls outside the edge. § Solutions? § Ignore the edges § The resultant image is smaller than the original § Pad with zeros § Introducing unwanted artifacts Digital Image Processing Lecture 5 8
![Values Outside the Range § Linear filtering might bring the intensity outside the display Values Outside the Range § Linear filtering might bring the intensity outside the display](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-9.jpg)
Values Outside the Range § Linear filtering might bring the intensity outside the display range. § Solutions? § Clip values § Scaling transformation New min New max Transform values in [g. L, g. H] to [0, 255] Digital Image Processing Lecture 5 9
![Separable Filters § Some filters can be implemented by successive application of two simpler Separable Filters § Some filters can be implemented by successive application of two simpler](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-10.jpg)
Separable Filters § Some filters can be implemented by successive application of two simpler filters. § Separability results in great time saving. § By how much? § For a mask of size nxn, for each image pixel § Originally, n 2 multiplications and n 2 -1 additions § After separation, 2 n multiplications and 2 n-2 additions Digital Image Processing Lecture 5 10
![Some terminology on Frequency… § Frequencies: a measure of the amount by which gray Some terminology on Frequency… § Frequencies: a measure of the amount by which gray](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-11.jpg)
Some terminology on Frequency… § Frequencies: a measure of the amount by which gray values change with distance. § High/low-frequency components: large/small changes in gray values over small distances. § High/low-pass filter: passing high/low components, and reducing or eliminating low/high-frequency filter. Digital Image Processing Lecture 5 11
![Low-Pass Filter § Mostly for noise reduction/removal and smoothing § 3 x 3 averaging Low-Pass Filter § Mostly for noise reduction/removal and smoothing § 3 x 3 averaging](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-12.jpg)
Low-Pass Filter § Mostly for noise reduction/removal and smoothing § 3 x 3 averaging filter to blur edges § Gaussian filter, § based on Gaussian probability distribution function § a popular filter for smoothing § more later when we discuss image restoration In 1 D: Digital Image Processing Lecture 5 12
![High-Pass Filter § The Laplacian, , is a high-pass filter, or § Sum of High-Pass Filter § The Laplacian, , is a high-pass filter, or § Sum of](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-13.jpg)
High-Pass Filter § The Laplacian, , is a high-pass filter, or § Sum of coefficients is zero § Insight: in areas where the gray values are similar (low-frequency), application of the filter makes the gray values close to zero. Digital Image Processing Lecture 5 13
![Laplacian Operator § Laplacian is a derivative operator – 2 nd order derivative § Laplacian Operator § Laplacian is a derivative operator – 2 nd order derivative §](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-14.jpg)
Laplacian Operator § Laplacian is a derivative operator – 2 nd order derivative § Highlighting graylevel discontinuities. § An edge detector § isotropic in x- and y-direction. Digital Image Processing Lecture 5 14
![Laplacian Operator (con’d) § We may add the diagonal terms too => isotropic results Laplacian Operator (con’d) § We may add the diagonal terms too => isotropic results](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-15.jpg)
Laplacian Operator (con’d) § We may add the diagonal terms too => isotropic results for increments of or § There are other variations/approximations: Separable filter Digital Image Processing Lecture 5 15
![Laplacian Filter: Demo Laplacian-filtered Original Digital Image Processing Lecture 5 16 Laplacian Filter: Demo Laplacian-filtered Original Digital Image Processing Lecture 5 16](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-16.jpg)
Laplacian Filter: Demo Laplacian-filtered Original Digital Image Processing Lecture 5 16
![Laplacian: Application § Laplacian highlights discontinuities, and turn featureless regions into dark background. § Laplacian: Application § Laplacian highlights discontinuities, and turn featureless regions into dark background. §](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-17.jpg)
Laplacian: Application § Laplacian highlights discontinuities, and turn featureless regions into dark background. § How can we make good use of the property? § One example is edge enhancement Laplacian enhanced original Digital Image Processing Lecture 5 17
![Edge Sharping § To make edges slightly sharper and crisper. § This operation is Edge Sharping § To make edges slightly sharper and crisper. § This operation is](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-18.jpg)
Edge Sharping § To make edges slightly sharper and crisper. § This operation is referred to as edge enhancement, edge crispening or unsharp masking. § A very popular practice in industry. § Subtracting a blurred version of an image from the image itself. Blurred image Sharpened image Original image Digital Image Processing Lecture 5 18
![Unsharp Masking: Why it works f(x, y) Digital Image Processing blurred image Lecture 5 Unsharp Masking: Why it works f(x, y) Digital Image Processing blurred image Lecture 5](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-19.jpg)
Unsharp Masking: Why it works f(x, y) Digital Image Processing blurred image Lecture 5 Sharpened 19
![Filter for Unsharp Masking § Combining filtering and subtracting in one filter. § Schematically, Filter for Unsharp Masking § Combining filtering and subtracting in one filter. § Schematically,](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-20.jpg)
Filter for Unsharp Masking § Combining filtering and subtracting in one filter. § Schematically, Subtract Original Blur with low-pass filter Digital Image Processing Scale with A<1 Lecture 5 Scale for display 20
![High-Boost Filtering § Generalization of unsharp masking § Here A is called boost coefficient, High-Boost Filtering § Generalization of unsharp masking § Here A is called boost coefficient,](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-21.jpg)
High-Boost Filtering § Generalization of unsharp masking § Here A is called boost coefficient, and § We rewrite the equation as § Applicable to any sharpening operation § can be § The filter for fhb becomes Digital Image Processing Lecture 5 21
![High-Boost Filtering: Demo § By varying A, better overall brightness can be improved. A=0 High-Boost Filtering: Demo § By varying A, better overall brightness can be improved. A=0](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-22.jpg)
High-Boost Filtering: Demo § By varying A, better overall brightness can be improved. A=0 original A=1. 7, much brighter A=1 Digital Image Processing Lecture 5 22
![Nonlinear Filters § Will discuss some of them in more detail later for the Nonlinear Filters § Will discuss some of them in more detail later for the](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-23.jpg)
Nonlinear Filters § Will discuss some of them in more detail later for the purpose of image restoration. § Maximum filter: the output the maximum value under the mask § Minimum filter: the output the minimum value under the mask § Rank-order filter: § Elements under the mask are ordered, and a particular value is returned. § Both maximum and minimum filter are instances of rank-order § Another popular instance is median filter Digital Image Processing Lecture 5 23
![More Nonlinear Filters § Alpha-trimmed mean filter: § Order the values under the mask More Nonlinear Filters § Alpha-trimmed mean filter: § Order the values under the mask](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-24.jpg)
More Nonlinear Filters § Alpha-trimmed mean filter: § Order the values under the mask § Trim off elements at either end of the ordered list § Take the mean of the remainder § E. g. assuming a 3 x 3 mask and the ordered list § Trimming of two elements at either end, the result of the filter is Digital Image Processing Lecture 5 24
![Summary § We introduced the concept of convolution § We briefly discussed spatial filters Summary § We introduced the concept of convolution § We briefly discussed spatial filters](http://slidetodoc.com/presentation_image/308f3ddcee315d0396f13c97437be33a/image-25.jpg)
Summary § We introduced the concept of convolution § We briefly discussed spatial filters of § Low-pass filter for smoothing § High-pass filter for edge sharpening § Nonlinear § We’ll come back to most of the filters under the appropriate topics. Digital Image Processing Lecture 5 25
- Slides: 25