Scribd
Upload
9 views

Image noise elimination algorithm

The document discusses image noise, its causes, types, and various elimination techniques, including spatial and frequency domain filtering methods. It highlights advanced noise reduction methods such as bilateral filtering, wavelet transform denoising, and deep learning-based approaches. Applications of these techniques span medical imaging, remote sensing, and computer vision, with future trends pointing towards AI-powered solutions.

Uploaded by

roopam12cse
Copyright
© © All Rights Reserved
Available Formats
Download as KEY, PDF, TXT or read online on Scribd
9 views

Image noise elimination algorithm

The document discusses image noise, its causes, types, and various elimination techniques, including spatial and frequency domain filtering methods. It highlights advanced noise reduction methods such as bilateral filtering, wavelet transform denoising, and deep learning-based approaches. Applications of these techniques span medical imaging, remote sensing, and computer vision, with future trends pointing towards AI-powered solutions.

Uploaded by

roopam12cse
Copyright
© © All Rights Reserved
Available Formats
Download as KEY, PDF, TXT or read online on Scribd
You are on page 1/ 13

PEC IT601D IMAGE PROCESSING CA1

IMAGE NOISE
ELIMINATION
ALGORITHM
NAME: ROOPAM BARMAN | ROLL NO: 34900122051 | REG NO: 223490110068
WHAT IS IMAGE PROCESSING?
Image noise refers to random variations in brightness or color information in digital images.
It is often introduced due to sensor limitations, transmission errors, or environmental factors.

Causes:
Sensor limitations
Electronic interference
Low light conditions
Transmission errors

Impact: Reduces clarity, affects analysis, and distorts important details.


TYPES OF IMAGE NOISE
1. Gaussian Noise
Characterized by a normal distribution of noise values.
Common in low-light and high-ISO images.
Formula:

Example: Night photography with grainy texture.
2. Salt-and-Pepper Noise
Appears as random black and white pixels in an image.
Usually caused by data corruption or transmission errors.
Easiest to remove using median filtering.
3. Poisson Noise (Shot Noise)
Due to photon fluctuations in low-light conditions.
More significant in scientific imaging and medical scans.
4. Speckle Noise
Grainy pattern seen in radar, ultrasound, and SAR imaging.
Caused by interference of coherent waves.
NOISE ELIMINATION TECHNIQUES
Spatial Domain Filtering
Directly modifies pixel values in an image.
Common techniques:
Mean Filtering (blurring effect)
Median Filtering (preserves edges)
Gaussian Smoothing (weighted averaging)
Adaptive Filtering (context-based filtering)
NOISE ELIMINATION TECHNIQUES
Mean Filtering
Concept:
Replaces a pixel value with the average of surrounding pixels.
Reduces Gaussian noise but causes blurring.
Mathematical Formula: 
Example:
A 3×3 filter kernel applied to an image.
NOISE ELIMINATION TECHNIQUES
Median Filtering
Concept:
Replaces each pixel value with the median of surrounding pixels.
Removes salt-and-pepper noise while preserving edges.
Algorithm:
Define a window (e.g., 3×3 kernel).
Sort pixel values within the window.
Replace the center pixel with the median value.
Application: Used in medical imaging, OCR, and object recognition.
NOISE ELIMINATION TECHNIQUES
Gaussian Filtering
Concept:
Uses a Gaussian function to apply a weighted average of neighboring pixels.
Smoother than mean filtering, better at noise reduction.
Gaussian Kernel (3×3 Example):

Application: Used in computer vision and preprocessing for edge detection.
FREQUENCY DOMAIN FILTERING
Fourier Transform for Noise Removal
Converts an image into the frequency domain.
Noise appears as high-frequency components.
Steps:
1 Compute Fourier Transform.
2 Apply low-pass filter to remove high frequencies.
3 Perform Inverse Fourier Transform to reconstruct image.
4

High-Pass vs Low-Pass Filtering


Low-Pass Filtering: Removes noise but causes blurring.
High-Pass Filtering: Enhances edges but retains noise.
Band-Pass Filters: Selectively remove specific noise frequencies.
ADVANCED NOISE REDUCTION METHODS
Bilateral Filtering
Concept:
Uses both spatial distance and pixel intensity difference.
Preserves edges while reducing noise.
Mathematical Formula

Application: Used in photography, medical imaging, and HDR processing.
Non-Local Means (NLM) Filtering
Compares patch similarity instead of just neighboring pixels.
Reduces Gaussian and Poisson noise effectively.
Computationally expensive but produces high-quality results.
ADVANCED NOISE REDUCTION METHODS
Wavelet Transform Denoising
Uses multi-resolution analysis to remove noise at different scales.
Steps:
1 Apply wavelet transform to decompose image.
2 Perform thresholding on high-frequency coefficients.
3 Reconstruct the image using inverse wavelet transform.
4

Application: Used in MRI and astronomical imaging.

Deep Learning-Based Denoising


Uses neural networks to learn noise patterns.
Popular models:
Denoising Autoencoders (DAEs)
Convolutional Neural Networks (CNNs)
Deep Image Prior (DIP)
Example: AI-based upscaling in Photoshop & smartphones.
PERFORMANCE METRICS FOR NOISE REDUCTION
How to Evaluate Denoising Algorithms?
Peak Signal-to-Noise Ratio (PSNR):

Structural Similarity Index (SSIM): Measures structural preservation.
Mean Squared Error (MSE): Measures average error.
APPLICATIONS OF NOISE ELIMINATION
Where Are These Algorithms Used?
1. Medical Imaging: MRI, CT scans.
2. Remote Sensing: Satellite and aerial imagery.
3. Computer Vision: Object detection, OCR.
4. Security & Surveillance: Enhancing low-light images
CONCLUSION & FUTURE TRENDS
Summary
Noise reduces quality, but various filtering techniques exist.
Traditional methods (Mean, Median, Gaussian, Fourier Transform) have limitations.
Deep learning models provide advanced noise removal.

Future Trends
AI-powered noise removal (GANs, Transformers).
Self-supervised learning for denoising.
Quantum image processing.

You might also like