Scribd
Upload
14 views4 pages

Case Study

The document discusses the application of wavelet transform for image denoising in digital image processing, specifically targeting noise removal while preserving key image features. It outlines the process of wavelet decomposition, thresholding high-frequency coefficients, and image reconstruction, highlighting the benefits of this method over traditional techniques. Wavelet-based denoising is particularly effective in medical imaging and other applications where detail preservation is crucial.

Uploaded by

vedantwadu111
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
14 views4 pages

Case Study

The document discusses the application of wavelet transform for image denoising in digital image processing, specifically targeting noise removal while preserving key image features. It outlines the process of wavelet decomposition, thresholding high-frequency coefficients, and image reconstruction, highlighting the benefits of this method over traditional techniques. Wavelet-based denoising is particularly effective in medical imaging and other applications where detail preservation is crucial.

Uploaded by

vedantwadu111
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 4

Vedant Wadu (221080073)

Batch-C
TY B.Tech

DIP - CASE STUDY


Problem Statement:
Image Denoising in Digital Image Processing (DIP) using Wavelet Transform
One of the most common challenges in digital image processing is noise removal. Noise can be
introduced during various stages of image acquisition, compression, or transmission. Noisy
images may contain pixel-level distortions that negatively affect image quality and make it
harder to analyze and interpret images. Typical examples of noise include Gaussian noise, salt-
and-pepper noise, and speckle noise.
A significant issue with traditional denoising techniques, such as Gaussian filtering, is that
while they effectively reduce noise, they tend to blur important features, especially edges and
details, which are crucial in applications such as medical imaging, satellite imaging, and
machine vision. Thus, the goal is to design a denoising method that removes noise effectively
while preserving the key image features such as edges and textures.
Problem: Given a noisy image (typically corrupted with Gaussian noise), the task is to
effectively denoise the image using wavelet transform techniques while maintaining its key
features (edges, textures, and details).

Solution: Using Wavelet Transform for Image Denoising


Wavelet transform is a powerful mathematical tool that enables the decomposition of an image
into various frequency components at multiple scales. It performs well in situations where
traditional methods struggle, particularly with noise removal in images, because it allows for
more sophisticated noise filtering while preserving important image details like edges.
Key Concepts of Wavelet Transform:
1. Wavelet Decomposition: The first step in wavelet-based denoising involves applying a
Discrete Wavelet Transform (DWT) to the noisy image. This process decomposes the
image into a set of subbands at different scales and orientations:
o Approximation coefficients (Low-frequency components): These represent the
smooth areas or global features of the image.
o Detail coefficients (High-frequency components): These represent the finer
details of the image, including edges and textures.
Vedant Wadu (221080073)
Batch-C
TY B.Tech

2. Thresholding: Once the image is decomposed into approximation and detail subbands,
noise is mostly concentrated in the high-frequency components. Therefore, the next step
is to threshold the high-frequency coefficients to remove the noise:
o Hard Thresholding: In this approach, if the wavelet coefficient is smaller than a
certain threshold, it is set to zero. This technique can be simple but might cause
abrupt changes, which could introduce artifacts in the reconstructed image.
o Soft Thresholding: In soft thresholding, the coefficients are shrunk toward zero by
a fixed amount, which smooths the result but avoids sharp changes. This is often
more effective in preserving the continuity of the image.
o Adaptive Thresholding: More sophisticated methods involve adjusting the
threshold for each coefficient depending on its location or its context in the image.
3. Wavelet Reconstruction: After thresholding the high-frequency components, the next
step is to reconstruct the image from the modified wavelet coefficients. The Inverse
Discrete Wavelet Transform (IDWT) is applied to the approximation and thresholded
detail coefficients to form the denoised image.
4. Fine-Tuning the Threshold: The thresholding value is one of the most critical factors in
the success of wavelet-based denoising. An inappropriate threshold value may either
leave too much noise or suppress important image features. Various strategies can be used
to select an optimal threshold, such as:
o Universal Thresholding: A simple method based on a mathematical formula that
provides a threshold value based on the noise level in the image.
o Bayesian Thresholding: A more advanced technique that considers both the noise
characteristics and the image content, providing a more adaptive and optimized
threshold value.
Step-by-Step Image Denoising Process Using Wavelet Transform:
1. Input: A noisy image is provided, where the noise is usually modeled as Gaussian noise
or salt-and-pepper noise.
2. Wavelet Decomposition:
o The image is decomposed using Discrete Wavelet Transform (DWT) into
approximation coefficients (low-frequency) and detailed coefficients (high-
frequency).
Vedant Wadu (221080073)
Batch-C
TY B.Tech

o For example, using Daubechies wavelets (db2, db4, etc.), which are commonly
used in image processing for their good balance of smoothness and compact
support.
3. Thresholding of High-Frequency Coefficients:
o Apply thresholding to the high-frequency coefficients to suppress the noise. The
threshold value can be determined through a fixed or adaptive strategy, depending
on the noise characteristics.
o Hard or soft thresholding methods are applied to remove coefficients that are likely
due to noise.
4. Inverse Wavelet Transform (IDWT):
o After thresholding, the image is reconstructed by applying the Inverse Discrete
Wavelet Transform (IDWT) to the modified coefficients (approximation and
detail).
o This produces the denoised image, which should exhibit less noise while retaining
important structures like edges.
5. Output: The output is a denoised image with less noise and preserved features, such as
edges and textures.

Benefits of Wavelet Transform for Image Denoising:


• Multiresolution Analysis: Wavelets allow decomposition of the image at multiple scales,
enabling noise removal at different frequencies. This is particularly beneficial for
capturing both global features (low-frequency components) and local features (high-
frequency components).
• Edge Preservation: Unlike traditional methods, wavelet-based denoising effectively
preserves edges and fine details. Since noise typically affects high-frequency
components, suppressing these components selectively allows for better preservation of
the essential details in the image.
• Efficient Noise Removal: Wavelet transform can more effectively identify and remove
noise from an image, especially when the noise is primarily concentrated in the high-
frequency subbands. This is in contrast to other techniques like Gaussian smoothing,
which blur edges and textures.
Vedant Wadu (221080073)
Batch-C
TY B.Tech

• Adaptability: Wavelet transforms can be customized to suit different types of noise (e.g.,
Gaussian, salt-and-pepper, speckle), and the thresholding step can be adjusted based on
the type of image and the characteristics of the noise.
• Robust to Different Noise Types: While Gaussian noise is commonly dealt with,
wavelet denoising can also be adapted to remove other types of noise, including impulse
noise or speckle noise, by modifying the thresholding function.

Example of Application:
Imagine a medical imaging scenario, where MRI scans or X-rays are taken with high noise due
to various factors (e.g., machine imperfections or transmission errors). Using wavelet-based
denoising, we can:
• Apply the DWT to decompose the noisy medical image.
• Threshold the high-frequency components to remove the noise.
• Reconstruct the denoised image while preserving crucial details like edges of anatomical
structures.
This technique ensures that the features needed for accurate diagnosis (such as the edges of
tumors or blood vessels) are not lost during the denoising process, unlike traditional methods
that might blur or smooth such important structures.

Conclusion:
Wavelet-based image denoising is an effective technique that allows for the removal of noise
from an image while preserving important structural features. By decomposing the image into
various frequency components and selectively thresholding the high-frequency components,
wavelet transforms offer a sophisticated method for noise reduction that traditional filtering
techniques cannot match. The multiresolution nature of wavelets provides a powerful tool for
tackling noise in various types of images, making this technique invaluable for a wide range of
image processing applications.

You might also like