Image

Enhancement
II
 Filtering
• It is applied for noise removal or to perform some
type of image enhancement
• Filtering can be done in the spatial domain as well
as frequency domain
• The operators we will discuss here are called spatial
filters since they operate on the raw image data in
the spatial domain
• They operate on the image data by considering small
neighborhoods in an image and returning the result
based on a linear or non-linear operations; moving
sequentially across and down the entire image
 Spatial Filtering
Spatial Filtering

Linear Filters Non-linear Filters

Average Filtering Smoothing Order-statistics

(Lowpass Filtering) Filters Filters
Weighted Averaging Sharpening
High-boost Filters Filters Median Filters
Derivative Filters
Smoothing filters: used for blurring and for noise reduction
Sharpening filters: used to highlight fine detail in an image or to
enhance detail that has been blurred, either in error or as a
natural effect of a particular method of image acquisition
 Linear Spatial Filtering

W (-1,-1) W (-1,0) W (-1,1)

Mask
coefficients,
W (0,-1) W (0,0) W (0,1) showing
Mask
coordinate
arrangement
W (1,-1) W (1,0) W (1,1)

Image f(x, y) f (x-1,y-1) f (x-1,y) f (x-1,y+1)

Pixels
Since the
of f (x,y-1) f (x,y) f (x,y+1) operations
image performed in the
section convolution are
under f (x+1,y-1) f (x+1,y) f (x+1,y+1) linear, these types
mask of filters are called
linear filters
 Linear Spatial Filtering
• The process consists simply of moving the filter
mask from point to point in an image
• At each point (x, y), the response of the filter at that
point is calculated using a predefined relationship
• The response is given by a sum of products of the
filter coefficients and the corresponding image
pixels in the area spanned by the filter mask
• The coefficient w (0,0) coincides with image value
f (x, y), indicating that the mask is centered at (x, y)
when the computation of the sum of products takes
place
 Linear Spatial Filtering
• For a 3x3 mask shown in figure, the response, R, of
linear filtering with the filter mask at a point (x, y)
in the image
R = w(-1,-1) f(x-1,y-1)+ w(-1,0) f(x-1,y)+……
+ w(0,0) f(x,y)+… + w(1,0) f(x+1,y)
+ w(1,1) f(x+1,y+1)
w(s,t) :(2a+1)×(2b+1) mask Where „a‟ & „b‟ are
nonnegative integers
• For a mask of size mxn, we assume that m = 2a+1 and
n = 2b+1
• Focus: mask of odd sizes, smallest meaningful size: 3x3
 Linear Spatial Filtering
• In general, Linear filtering of an image f of size MxN
with a filter mask of size mxn is given by:
a b
g ( x, y )    w(s, t ) f ( x + s, y + t )
s   at   b

where a = (m -1)/2 and b = (n -1)/2

• To generate a complete filtered image this equation must be
applied for x = 0, 1, 2,…, M-1 and y = 0, 1, 2,…,N-1
– In this way, we are assured that the mask processes all pixels
in the image
• Nonlinear spatial filters also operate on neighborhoods, and the
mechanics of sliding a mask past an image are the same. They do
not explicitly use coefficients in the sum-of-products manner.
 Linear Spatial Filtering
• Border Effects- Special Consideration
– For a small value of m and n, it is not that much
problem; otherwise we may loose substantial amount of
data
– Common procedures:
• Border strips are added and set to zero
• Rows and columns are considered to wrap around
• Limit the excursion of the center of the filter mask to a
distance no less than (n-1)/2 pixels from the border of
the original image
– The resulting filtered image will be smaller than the
original, but all the pixels in the filtered image will
have been processed with the full mask
 Spatial Correlation and Convolution

• Correlation is the process of moving a filter mask

over the image and computing the sum of
products at each location, as explained

• The mechanics of convolution are the same,

except that the filter is first rotated by 1800
Spatial Correlation and Convolution
Spatial Correlation and Convolution

• See that convolution of

a function with an
impulse copies the
function at the location
of the impulse
• If the filter mask is
symmetric, correlation
and convolution yield
the same result
 Spatial Correlation and Convolution

• Correlation a b
w( x, y ) f ( x, y )    w(s, t ) f ( x + s, y + t )
s   at   b
• Convolution
a b
w( x, y ) f ( x, y )    w(s, t ) f ( x  s, y  t )
s   at   b

where a = (m -1)/2 and b = (n -1)/2

Minus sign denotes 1800 rotation of f, which has been

done for notational simplicity and to follow convention
 Spatial Correlation and Convolution
• Using correlation or convolution to perform spatial
filtering is a matter of preference
– Either can perform the function of the other by a simple
rotation of the filter
– Important is the filter mask used in a given filtering task be
specified in a way that corresponds to the intended operation
• The linear spatial filtering discussed here are based
on correlation
• Convolution filter, Convolution mask, Convolution
kernel  spatial filter (not necessarily used for true convolution)
• Convolving a mask with an image is often used to
denote the sliding, sum-of-products process
Linear Filtering: Vector Representation

• It is convenient to write the response R of

an m x n mask at any point (x, y) is:
mn
R  w1 z1 + w2 z 2 + ... + wmn z mn   wi zi
i i

ws are the coefficients of an m x n W1 W2 W3

filter and the zs are the corresponding
W4 W5 W6
image intensities encompassed by
the filter W7 W8 W9
 Generating Spatial Filter Mask
• Mask Coefficient
– Proper selection of the coefficients and application of
the mask at each pixel position in an image makes
possible a variety of useful image operations
• Noise reduction
• Region thinning
• Edge detection
– Coefficients are selected based on what the filter is
supposed to do, keeping in mind that all we can do with
linear filtering is to implement a sum of products
• Applying a mask at each pixel location in an image
is a computationally extensive task
 Generating Spatial Filter Mask
• Example:
0 1 0
Consider the image: 1 1 1
0 0 1
Suppose the second row of the image is to be
highlighted, what is the mask that is needed?
 Generating Spatial Filter Mask
• One interesting observation:
– Overall effect can be predicted on the general pattern, if
one uses convolution mask
• If the sum of coefficients of the mask is one,
average brightness of the image will be retained
• If the sum of coefficients of the mask is zero,
average brightness of the image will be lost and
will return a dark image
• If coefficients are alternating positive and negative,
the mask is a filter that will sharpen an image
• If coefficients are all positive, it is a filter that will
blur the image
 Smoothing Spatial Filters
• Used for blurring and for noise reduction
• Blurring is used in preprocessing steps, such as
– Removal of small details from an image prior to large
object extraction, and
– Bridging of small gaps in lines or curves
• Noise reduction can be accomplished by blurring
with a linear filter and also by nonlinear filtering
• The output/response of a smoothing, linear spatial
filter is simply the average of the pixels contained
in the neighborhood of the filter mask
• Also called „Averaging Filters‟ or „Lowpass
Filters‟
 Smoothing Linear Filters
• Replacing the value of every pixel in an image by the
average of the gray levels in the neighborhood will reduce
the “sharp” transitions in gray levels
• Sharp intensity transitions
– random noise in the image
– edges of objects in the image
• Thus, smoothing can reduce noises (desirable) and blur
edges (undesirable)
• Other uses
– smoothing of false contours resulting from using an insufficient
number of intensity levels
– reduction of irrelevant detail in an image (here, “irrelevant”
means pixel regions that are small wrt the size of filter mask)
 General Form : Smoothing Mask
• Filter of size m x n (m and n odd)
a b

  w(s, t ) f ( x + s, y + t )
g ( x, y )  s   at   b
a b

  w(s, t )
s   at   b

summation of all coefficient of the mask (computed once)

The complete filtered image is obtained by applying the operation

for x = 0, 1, 2, …., M – 1 and y = 0, 1, 2, …., N - 1
3x3 Smoothing Linear Filters

box filter weighted average

the center is the most important and

other pixels are inversely weighted as
a function of their distance from the
center of the mask
 Weighted Average Filter
• The basic strategy behind weighting the center
point the highest and then reducing the value of
the coefficients as a function of increasing
distance from the origin is simply an attempt to
reduce blurring in the smoothing process.

Weighted Average
Filter
Example 1
(a) Original image of size
a b 500x500 pixel
c d (b)-(f) Results of smoothing with
e f square averaging filter masks of
sizes n = 3, 5, 9, 15, and 35
respectively
Note:
- details that are of approx. the
same size as the filter mask are
affected considerably more
- the size of the mask establishes
the relative size of the objects
that will be blended with the
background
- big mask is used to eliminate
small objects from an image