Chapter-2 Image Enhancements
What is Image Enhancement?
Image Enhancement refers to the process of improving the visual appearance
of an image or making its features more distinguishable for analysis.
It involves modifying pixel intensity values to highlight certain characteristics,
such as edges, contrast, or brightness.
The goal of enhancement is to make an image more suitable for human
observation or to prepare it for further image processing tasks (e.g., object
recognition).
Image Enhancement in Two Domains
Image enhancement can be performed in two domains:
1. Spatial Domain
2. Frequency Domain
1. Spatial Domain
In the spatial domain, image enhancement is performed directly on the pixels
of the image.
Operations include pixel intensity transformation, manipulation, or filtering
techniques applied on the pixel values of an image.
These operations work by altering the intensity values (gray levels) of pixels
and are computationally simpler than frequency domain techniques.
2. Frequency Domain
In the frequency domain, enhancement is done by transforming the image
into a frequency space using techniques like the Fourier Transform.
This allows the modification of specific frequency components, enabling more
targeted enhancement of image features such as noise reduction or
sharpening edges.
Chapter-2 Image Enhancements 1
What is the Spatial Domain?
Spatial domain refers to the actual space in which the image exists—
essentially, the image is treated as a matrix of intensity values corresponding
to pixel locations.
In this domain, operations are performed directly on the pixel values to
enhance or extract certain image features.
Mathematical Representation:
Spatial Domain Process
The spatial domain process involves applying a transformation function \( T \)
directly to the pixel values of an image. Common techniques in this domain
include:
1. Gray-Level Transformations
Directly modify pixel intensity values to improve contrast or brightness.
Examples: Contrast stretching, logarithmic transformations, and power-law
transformations.
2. Image Smoothing
Smoothing filters are applied to reduce noise and make the image less sharp
by averaging the intensity values of pixels within a neighborhood.
Examples: Mean filter, Gaussian filter.
3. Image Sharpening
Chapter-2 Image Enhancements 2
Sharpening filters enhance edges by accentuating the differences in pixel
intensities.
Examples: Laplacian filter, gradient operators (Sobel, Prewitt).
4. Spatial Filtering
Applies convolution with a kernel or mask to each pixel and its neighbors,
where the kernel represents the desired operation (such as smoothing or
sharpening).
Examples: Median filtering, high-pass filtering.
Point Processing
Point processing is a type of image enhancement technique where the
transformation applied to each pixel depends only on that pixel's value. In this
approach, no neighborhood or surrounding pixel information is considered.
This is the simplest form of spatial domain processing because it manipulates
each pixel individually without reference to its neighbors.
Mask Processing (Neighborhood Processing)
Mask processing (or filtering) is a technique where the transformation applied
to a pixel depends on the pixel's value as well as the values of its neighboring
pixels. This process involves a mask or kernel that is applied to each pixel and
Chapter-2 Image Enhancements 3
its neighborhood to perform operations such as blurring, sharpening, and
edge detection.
Some Basic Gray Level Transformations
Chapter-2 Image Enhancements 4
Histogram Processing
What is a Histogram?
Chapter-2 Image Enhancements 5
In image processing, a histogram is a graphical representation of the distribution
of pixel intensities (gray levels) in an image. It displays how often each intensity
value occurs, providing insight into the image's contrast, brightness, and dynamic
range.
Discrete Function - \( h(r_k) = n_k \)
Example - 4 x 4, 3-bit Gray Scale Image
Let's consider a 4x4 grayscale image with 3-bit depth. This means the gray levels
range from 0 to 7 (since 3 bits can represent 8 different values).
Assume the following is the pixel intensity matrix of the image:
2 2 3 4
1 1 2 3
4 5 6 7
7 7 7 6
To construct the histogram:
1. Count the occurrences of each gray level:
Gray Level 0: 0 occurrences
Gray Level 1: 2 occurrences
Gray Level 2: 4 occurrences
Gray Level 3: 3 occurrences
Gray Level 4: 2 occurrences
Gray Level 5: 1 occurrence
Chapter-2 Image Enhancements 6
Gray Level 6: 2 occurrences
Gray Level 7: 4 occurrences
2. Plot the histogram:
Gray Level Frequency
0 0
1 2
2 4
3 3
4 2
5 1
6 2
7 4
You can represent this information visually in a bar chart:
Frequency
5| X
4| X X
3| X X X
2| X X X X
1| X X X XX
0 |----------------
0 1 2 3 4 5 6 7
Gray Level
In this bar chart:
The x-axis represents the gray levels (0 to 7).
The y-axis represents the frequency (number of occurrences) of each gray
level in the image.
Histogram Processing
Chapter-2 Image Enhancements 7
Histogram processing involves manipulating the histogram to improve or alter the
image. Common techniques include:
Histogram Equalization: Enhances contrast by spreading out the most
frequent intensity values.
Histogram Stretching: Expands the range of pixel values to utilize the full
dynamic range of the image.
Histogram Matching: Adjusts the histogram of an image to match a reference
histogram.
These techniques help in enhancing image quality and making features more
distinguishable.
Four basic Image types & their Histograms
Normalized Histogram
What is it?
Chapter-2 Image Enhancements 8
A normalized histogram represents the relative frequency of each gray level in an
image rather than the absolute count. It is obtained by dividing the frequency of
each gray level by the total number of pixels in the image.
Why Normalize the Histogram?
1. Comparison Across Images: Normalization allows for comparison of
histograms across images of different sizes or pixel counts because it scales
the frequencies to a common range (0 to 1).
2. Image Processing Techniques: Some image processing techniques and
algorithms require normalized histograms, such as those used in histogram
equalization, where the goal is to spread out pixel values evenly.
3. Probabilistic Interpretation: The normalized histogram provides a probability
distribution of gray levels, which can be useful in statistical image analysis and
pattern recognition.
Example Calculations
For a 3 x 3 Black Image
Consider a 3 x 3 image where all pixels are black (gray level 0):
000
000
000
Chapter-2 Image Enhancements 9
For a 10 x 10 Black Image
Consider a 10 x 10 image where all pixels are black (gray level 0):
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
0000000000
Histogram Processing
Chapter-2 Image Enhancements 10
Chapter-2 Image Enhancements 11