Spatial Domain Filtering and

Intensity Transformations

Spatial domain is simply the plain containing

pixels of image.

Spatial domain techniques are more efficient

computationally and requires less
processing.

Spatial domain refers to aggregate of pixels

that compose an image.
The spatial domain process is explained by
following expression.

g(x,y) = T[f(x,y)]

Where f(x,y) is the input image and g(x,y) is

the image at output.

T is an operator on f(x,y) defined over

neighborhood of point (x,y).

The image shows one such transformation.

An example of such transformation is image

averaging.
 A 3x3 neighbourhood about
(x,y)
The square of size 3x3 will start from origin
and will scan whole image horizontally and
vertically.
There is no neighborhood pixels at border
hence 0 or some other intensity value is
assumed.

The process just described is called as

Spatial Filtering.

The smallest neighborhood possible is 1x1 in

which new value of pixel depends on itself
and in this case T is called as Intensity
Transformation Function.

Basic intensity transformation functions

include image negatives, log transform etc.
Example of contrast stretching and
thresholding
In image intensity transformation the relation
between input and output is given by

s= T(r)

r is pixel before processing and s is output pixel

value.

A. Image Negatives

The relation between input and output is given by

s=L-1-r

It produces equivalent of image negative.

It is used for enhancing white/gray details in dark

B. Log Transformation

s= clog(1+r)

Where c is constant and r >=0.

It maps narrow range of low level in wide

range of output levels and vice versa.

It expands dark pixel value while

compressing higher pixel values.

Inverse log transforms is also there which

does the opposite task of log transformation.

Example in Fourier transform.

C. Power Law Transformation

As name suggests, it has exponential relation

between input and output.

s = crγ

Or some times as s = c(r+eps)γ

It is better than log transform as more

possibilities at outputcorrection
The term gamma are present depending
used in TV
on value of
industry [Link] law transformation.
performs
 D. Piecewise Linear Transformation
Piecewise Linear transformation functions are
less complex than other functions.

In many case piecewise linear implementation

is the more practical approach.

The disadvantage is that it requires more

input from user.

Contrast stretching transformation can be

implemented as piecewise linear
approximation.

The curve shape is controlled by points (r1,s1)

Consider the piecewise linear function for
following three cases

1.r1 = s1 and r2 = s2

2.r1 = r2 and s1 = 0 and s2 = L-1

3.r1<r2 and s1<s2

Intensity Level Slicing

Slicing is done to highlight certain range of
intensities in an image. This is done by two
approaches
The first approach in intensity slicing could be
to display range of interest in one colour and
all other intensity by other.

In second approach colour of range of

intensities is changed and only and other
intensities remain as it is
Example
Bit Plane Slicing

We can make a binary image by using singke

bit of every pixel.

If L=256 every pixel would contain 8 bits thus

we can make 8 binary images starting from
LSB to MSB.

By doing this we can highlight contribution of

specific bits in an image.
Histogram Processing

The histogram of a digital image with

intensity levels in intensity range [0,L-1]is a
discrete function h(rk) = nk where rk is the kth
intensity value and and nk is the number of
pixels in the image with intensity rk.

Histogram can be normalized by dividing

each of its components by total number of
pixels present in the image.

If MN is total number of pixels where M is

the number of rows in the image and N is the
coloumns.
Then normalized histogram is given by:

p(rk) = nk/MN

It can be said that p(rk) is the estimate of the

probability of occurrence of intensity level
rk.

Histograms forms the basis of numerous

spatial domain processing techniques.

Histogram can be used for image

enhancements also histogram equalization is
one of an example of image enhancement by
Histogram Equalization

Let the variable r represent the gray levels of

the image to be enhanced.

We assume that the transformation function

T(r) satisfies the following conditions:

(a)T(r) is single-valued and monotonically

increasing in the interval 0<= r <= 1; and

(b) 0 <=T(r) <=1 for 0 <=r <=1.

Let us discuss the histogram Equalization in
detail.

First we shall discuss it for continuous pdf

and then we will extend it to discrete pdf
case.
Example

Consider 64x64
pixel 3bit image
L=8 and L-1 =7
with specification
in adjoining table
Histogram Matching

Histogram equalization process automatically

determines the function which generates an
image with uniform histogram at the output.

In many cases the uniform histogram is not

the required output and we want a histogram
to acquire a specific shape or distribution at
the output.

The method used to generate a processed

image that has a specified histogram is called
histogram matching or histogram
specification.
Let us consider continuous gray levels r and
z, and let pr(r) and pz(z) denote their
corresponding continuous probability density
functions.

Where, r and z denote the gray levels of the

input and output (processed) images,
respectively.

We can estimate pr(r) from the given input

image, while pz(z) is the specified probability
density function that we wish the output
image to have.

We will first determine random variable s

Fundamentals of Spatial Filtering
Spatial filtering is an important tool in image
processing and caters to broad range of
applications.

Spatial filter has 2 characteristics:

[Link]
2.A predefined operation, that generates the
new pixel value.

The coordinates of new pixel is same as the

center of neighbourhood.

At each point (x, y), the response of the filter at

that point i.e. g(x,y) is the sum of products of
filter coefficients and the image pixels
g(x,y) = w(-1,-1)*f(x-1,y-1) + w(-1,0)*f(x-1,y) +
…w(0,0)*f(x,y) + …. w(1,1)*f(x+1,y+1)
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.

For a mask of size m*n, it is assumed that

m=2a+1 and n=2b+1, where a and b are
nonnegative integers.

It means that masks are of odd sizes, with the

smallest meaningful size being 3*3.
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.

The convolution is the same but filter is first

rotated by 180 degrees.

The figure shows both the processes on 1 D

data.

The data is first padded by m-1 zer0s on both

sides where m is the size of filter.

The same process could be extended to 2D data

2D Correlation

2D Convolution

a =? b = ?
Smoothing Spatial
Filters
Smoothing filters are used for blurring and
noise reduction.

The implemented filter could be linear or non

linear.

The linear filter is one in which the relationship

between input and output is linear.

Similarly, non linear filters can be implemented.

Example of linear filters is: Averaging Filter

Example of Non - linear filters is: Median

Where blurring is implemented in
preprocessing tasks to remove small details
from an image prior to large object extraction.

The output of a smoothing (averaging or

lowpass) linear spatial filter is the average of
the pixels contained in the neighborhood of the
filter mask.

By replacing the value of every pixel in an

image by the average of the intensity levels in
the neighborhood defined by a filter mask, the
resulting image will have reduced “sharp”
transitions in intensities.
As random noise typically corresponds to such
transitions, we can achieve denoising.

However, edges are often characterized by

sharp intensity transitions, so smoothing linear
filters may have the undesirable side effect on
edges.

Examples of such masks:

1) A box filter – spatial averaging filter 3x3
2) Weighted average filter attempt to reduce
blurring
The second mask, shown in Figure is called
weight average, thus giving more importance
(weight) to some pixels at the expense of others.

The general implementation for filtering an M

×N image with a weighted averaging filter of
size m ×n is given by the expression
Order-statistic (nonlinear) filters

Order-statistic filter are nonlinear spatial

filters.

There response is based on ordering (Ranking)

the pixels in the neighborhood and then
replacing the value of the center pixel by the
value determined by the ranking result.

The median filters are quite effective against

the impulse noise (salt-and-pepper noise).

Ex: the 3x3 neighborhood has values (10, 20,

20, 20,15, 20, 100, 25, 20). These values are
Sharpening Spatial Filters
Sharpening highlight transitions in intensity.

It is used in areas like electronic printing,

medical imaging, industry and military
applications.

Image smoothing requires blurring of images

which performs averaging operation which is
analogous to integration.

So, sharpening operation can be accomplished

by spatial differentiation.

The image differentiation enhances edges and

other discontinuities and deemphasizes areas
First and Second Order Derivative
The derivatives of a digital function are defined
in terms of differences. There are various ways
to define these differences.

Following are the condition that a first order

derivative must fulfill:
a. must be zero in flat segments (areas of constant
gray-level values)
b. must be nonzero at the onset of a gray-level step or
ramp and
c. must be nonzero along ramps.

Similarly a second order derivative must fulfill

following condition.
a. must be zero in flat areas
b. must be nonzero at the onset and end of a gray-level
step or ramp
A basic definition of the first-order derivative of a one-
dimensional function f(x) is the difference given by:

Similarly, a second-order derivative can be defined as

the difference
First-order derivatives generally produce
thicker edges in an image.

Second-order derivatives have a stronger

response to fine detail, such as thin lines and
isolated points.

First order derivatives generally have a stronger

response to a gray-level step.

Second order derivatives produce a double

response at step changes in gray level.

In most applications, the second derivative is

better suited than the first derivative for image
enhancement because of its ability to enhance
The Laplacian
The Laplacian is a two-dimensional, second
order derivatives for image enhancement.

The Laplacian is an isotropic derivative

operator , which, for a function (image) f(x, y) of
two variables, is defined as:

The filter mask created by Laplacian generates

isotropic filters, which means their response is
independent of the direction of the
discontinuities in the image.

The Laplacian are the simplest isotropic

derivative operator. The word isotropic denotes