Department Of Computer

Engineering

BE MINI PROJECT LP-VI(NLP)

"Feature Extraction using Zernike

Presented by :
Aishwarya Yogesh Chandwadkar (BE-A-30)
Atharva Nitin Chandwadkar (BE-A-31)
Kalyani Yogesh Desai (BE-A-39)
Sarvesh Uday Joshi (BE-A-71)

Guided By-
Prof. A. P. Shiralkar
Aim: Feature Extraction using Zernike Moments

Objective: To extract the features using the Zernike Moments

Software & Hardware Requirements:

● PC/Laptop
● Windows
● Python Libraries
● Google Colaboratory

Scope:

Feature extraction plays a crucial role in various fields such as computer vision and image
processing. One powerful technique for extracting features from images is Zernike moments.
Zernike moments are a set of orthogonal polynomials defined over a unit circle, which can be
used to represent shape and texture information in an image. In this article, we will delve into
the concept of feature extraction using Zernike moments and explore its applications.

Zernike moments are calculated by integrating the product of the image intensity and the Zernike
polynomials over the region of interest. These polynomials possess desirable properties, such as
being orthogonal and rotationally invariant, making them suitable for feature extraction tasks. By
calculating Zernike moments up to a certain order, we can obtain a set of coefficients that capture
the image's shape and texture characteristics.

One of the significant advantages of using Zernike moments for feature extraction is their
ability to represent complex shapes and contours accurately. Unlike other feature extraction
methods, Zernike moments can capture details at multiple scales, allowing for a more
comprehensive representation of an object's shape. This makes them particularly useful in
applications such as object recognition, where robust shape representation is essential.
Flowchart :

FIG: Process
Algorithm:

Step 1) :- Start

Step 2) :- Load an image for feature extraction

Step 3) :- Preprocess the image if necessary (e.g., convert to grayscale, resize, denoise)

Step 4) :- Select the region of interest (ROI) within the image

Step 5) :- Set the maximum order of Zernike moments to be calculated

Step 6) :- For each order 'n' from 0 to the maximum order:

a. For each repetition 'm' from -n to n:

i. Initialize the Zernike moment coefficient (Zn,m) to 0

ii. Calculate the normalization factor (Cn,m) using the formula:

Cn,m = sqrt((2n + 2) / (pi * (1 + delta(m, 0))))

iii. For each pixel within the ROI:

A. Convert the pixel coordinates to polar coordinates (radius, angle)

B. Calculate the Zernike polynomial value (Rn,m) using the radial polynomial and azimuthal
function

C. Update the Zernike moment coefficient:

Zn,m = Zn,m + (image_intensity * Rn,m * Cn,m)

Step 7) :- Store the Zernike moment coefficients (Zn,m) as a feature vector.

Step 8) :- Perform further analysis or application-specific tasks using the feature vector (e.g.,
object recognition, texture classification, medical image analysis.

Step 9) :- End
Theory:

Feature extraction plays a crucial role in various fields such as computer vision and image
processing. One powerful technique for extracting features from images is Zernike moments.
Zernike moments are a set of orthogonal polynomials defined over a unit circle, which can be
used to represent shape and texture information in an image. In this article, we will delve into
the concept of feature extraction using Zernike moments and explore its applications.

Zernike moments are calculated by integrating the product of the image intensity and the Zernike
polynomials over the region of interest. These polynomials possess desirable properties, such as
being orthogonal and rotationally invariant, making them suitable for feature extraction tasks. By
calculating Zernike moments up to a certain order, we can obtain a set of coefficients that capture
the image's shape and texture characteristics.

One of the significant advantages of using Zernike moments for feature extraction is their
ability to represent complex shapes and contours accurately. Unlike other feature extraction
methods, Zernike moments can capture details at multiple scales, allowing for a more
comprehensive representation of an object's shape. This makes them particularly useful in
applications such as object recognition, where robust shape representation is essential.

Furthermore, Zernike moments have proven to be robust to noise and image transformations,
making them suitable for real-world scenarios. Their rotational invariance property ensures that
the extracted features remain consistent even when the object is rotated in the image. This
characteristic is particularly valuable in tasks like face recognition, where facial features need
to be recognized irrespective of the subject's orientation.

Another advantage of Zernike moments is their compactness. The set of coefficients obtained
from Zernike moment calculations can be considered as a feature vector that represents the
image. These feature vectors are typically of much lower dimension compared to the original
image, reducing storage and computational requirements. This compact representation makes
Zernike moments suitable for applications with limited resources, such as mobile devices and
embedded systems.

The applications of Zernike moments extend beyond shape representation and object
recognition. They have found utility in various fields, including medical image analysis, texture
classification, and image retrieval. In medical imaging, Zernike moments can be used to
characterize anatomical structures, detect abnormalities, and aid in diagnosis. In texture
classification, Zernike moments can capture textural patterns and discriminate between different
material surfaces. In image retrieval, Zernike moments can be used to match and retrieve similar
images based on their shape and texture features.

So Basically, feature extraction using Zernike moments is a powerful technique for representing
shape and texture information in images. With their ability to accurately capture complex shapes,
rotational invariance, and compact representation, Zernike moments have proven to be valuable
in a wide range of applications. As technology advances, further research and refinement of
Zernike moment-based feature extraction methods are likely to unlock new possibilities in image
analysis, computer vision, and beyond.
Code & Output:
Conclusion:

In conclusion, the feature extraction technique using Zernike moments offers valuable
advantages for image analysis and computer vision applications. By capturing shape and texture
information, Zernike moments provide a comprehensive representation of complex contours and
objects, making them suitable for tasks such as object recognition. The rotational invariance
property ensures consistent feature extraction regardless of the object's orientation, making it
particularly useful in face recognition and other similar applications. The compactness of the
resulting feature vectors reduces storage and computational requirements, making Zernike
moments suitable for resource-constrained systems.