A Mini Project Report on

“Image Denoising Using Autoencoder”

Submitted in partial f u l f i l l m e n t of the requirements for the award of the degree of

Bachelor of Engineering

in

Computer Science & Engineering (AI & ML) Department

by

03. Prachi Bhoir

06. Anand Borkar
19. Hemant Kadam
27. Onkar Lohar

Under the Guidance of

Prof. Ashwini Garole

Vishwaniketan Institute of Management Entrepreneurship and Engineering Technology

Computer Science & Engineering (AI & ML) Department

UNIVERSITY OF MUMBAI

Academic Year 2023-2024

 Vishwaniketan Institute of Management
Entrepreneurship and Engineering Technology

Computer Science & Engineering (AI & ML) Department

CERTIFICATE

This is to certify that the project entitled “Image Denoising using

Autoencoder” submitted by “Prachi Bhoir, Anand Borkar, Hemant Kadam,
Onkar Lohar” for the partial fulfilment of the requirement for the award of a
degree Bachelor of Engineering in CSE (AIML) to the University of Mumbai, is
a bonafide work carried out during the academic year 2023-2024

Prof. Ashwini Garole Dr. Ankush Pawar Dr. B R Patil

Guide Head CSE(AIML) Principal
CSE(AIML) Department VIMEET, Khalapur
Department VIMEET, Khalapur

Place: ViMEET, Khalapur

Date:
 Acknowledgement

We have great pleasure in presenting the report on Image denoising using Autoencoder.
We take this opportunity to express our sincere thanks towards my guide Prof. Ashwini
Garole Department of CSE(AIML), ViMEET, Khalapur for providing the technical
guidelines and suggestions regarding the line of work. We would like to express my gratitude
for his/her constant encouragement, support and guidance through the development of the
project.
We thank Dr. Ankush Pawar Head CSE(AIML), ViMEET for his encouragement
during the progress meeting and for providing guidelines to write this report.
We thank Prof. Ashwini Garole mini project coordinator, CSE(AIML), ViMEET for
being encouraging throughout the course and for guidance.
We also thank the entire staff of ViMEET for their invaluable help rendered during the course
of this work. We wish to express our deep gratitude to all my colleagues at ViMEET for their
encouragement.

Prachi Bhoir Anand Borkar Hemant Kadam Onkar Lohar

 Index
TITLE PAGE NO.
List of Figures 5
List of Tables 5
Abstract 6
Contents
Chapter 1 1 Introduction 7
Chapter 2 2 Literature Review 8
2.1 Motivation 9
2.2 Problem Definition 9
2.3 Aim 9
2.4 Objectives 10
Chapter 3 3 Proposed System 11
3.1 Working 12
3.2 Details of Hardware & Software 13
3.3 Methodology 14
Chapter 4 4 Code 15
Chapter 5 5 Result 16
Chapter 6 6 Conclusions and Future Scope 17
Chapter 7 7 References 18
 List of Figures

Figure Page
Description
No. No.

3.1 Flowchart 11

List of tables

Table No. Description Page No.

2.1 Literature Review

 ABSTRACT
An autoencoder is a type of artificial neural network used for unsupervised learning of efficient
coding. They aim to learn a representation (encoding) for a set of data, typically for the purpose of
dimensionality reduction.
For image denoising, an autoencoder can be trained on clean images, and then it can be used to
remove noise from noisy images by feeding the noisy images as input and obtaining the denoised
images as output.
Image Denoising: Use an autoencoder to remove noise from images. Train the autoencoder on
images with added noise and then use it to reconstruct the original, noise-free image.
 Chapter 1
Introduction
Image denoising is the process of removing noise from digital photographs while retaining
important visual components. Addition of noise during the process of taking, sending, or
processing images might reduce their quality and complicate their accurate evaluation or
analysis.
Image denoising is essential in many industries, such as surveillance, photography, medical
imaging, and satellite imaging. It enhances visual interpretation, image quality, and the
effectiveness of subsequent image processing tasks (such as object identification or
recognition).
The autoencoder neural network design is a widely used kind for image denoising. They
consist of two main parts: an encoder and a decoder. Once the encoder has compressed the
data, the decoder reconstructs the denoised image.

A denoising autoencoder is a type of neural network architecture designed to remove noise

from input data. It is based on the concept of autoencoders, which are neural networks used
for learning efficient representations of input data. However, denoising autoencoders are
specifically trained to reconstruct clean versions of corrupted or noisy input data.

Working:

Architecture: Similar to traditional autoencoders, a denoising autoencoder consists of an

encoder and a decoder. The encoder compresses the input data into a lower-dimensional
representation (latent space), while the decoder reconstructs the original input data from this
representation.

Corrupted Input: In denoising autoencoders, the input data is intentionally corrupted by

adding noise. This noise can be random or follow a specific distribution, such as Gaussian
noise.

Training Objective: The denoising autoencoder is trained to minimize the reconstruction error
between the clean input data and the denoised output data. During training, the model learns
to map noisy input data to clean output data, effectively removing the added noise in the
process.

Noise Removal: By learning to reconstruct clean versions of noisy input data, the denoising
autoencoder implicitly learns to filter out the noise while preserving the important features of
the input data.

Applications: Denoising autoencoders have various applications, including image denoising,

signal denoising, and anomaly detection. In image denoising tasks, denoising autoencoders
are particularly effective at removing noise from images while retaining important visual
details.
 Chapter 2
Literature Review
Sr. Paper Title Author Key points Limitations
No

1 Denoising Vincent et al. Introduced denoising Limited to

Autoencoders autoencoders (DAEs) for grayscale images,
learning robust features may not scale well
by reconstructing clean to high-resolution
inputs from noisy inputs. or colour images.

2 Beyond a Zhang et al. Proposed residual May require large

Gaussian learning of deep CNNs amounts of
Denoiser: for image denoising, labelled data for
Residual achieving state-of-the-art training.
Learning of performance on
Deep CNN for benchmark datasets.
Image
Denoising

3 Learning to Wang et al. Investigated learning to Performance may

Denoise denoise images with deep degrade with
Images with CNNs, achieving complex noise
Deep CNNs competitive denoising patterns or low
performance on signal-to-noise
benchmark datasets. ratios.

4 Deep Chen et al. Reviewed deep learning Lack of

Learning in techniques for image standardized
Image denoising, covering benchmarks and
Denoising: various models, evaluation metrics
From Models architectures, and may hinder
to Algorithms optimization strategies comparison
used in recent research. between methods.

Table 2.1: Literature Review

2.1 Motivation
1 Practical Significance:
Image denoising is a crucial preprocessing step in various fields such as medical
imaging, surveillance, remote sensing, and digital photography. Clean images are
essential for accurate analysis, diagnosis, and decision-making in these domains.
Developing an effective denoising model can significantly improve the quality and
reliability of image data in practical applications.

2 Challenges in Noise Removal:

Traditional image denoising methods often struggle to effectively remove noise while
preserving important image details. Autoencoder-based approaches offer a promising
solution by leveraging the representation learning capabilities of neural networks to
extract meaningful features and reconstruct clean images from noisy inputs.

3 Research Gap and Innovation:

While autoencoder-based image denoising has shown promising results, there are still
challenges and opportunities for improvement. This project seeks to address these
gaps by investigating novel architectures, optimization strategies, and evaluation
metrics for image denoising. By contributing new insights and innovations to the
field, the project aims to advance the state-of-the-art in image processing.

4 Educational and Learning Goals:

Undertaking this project provides an excellent opportunity for hands-on learning and
skill development in deep learning, image processing, and algorithm development.
Through practical experimentation and research, participants can deepen their
understanding of neural networks and their applications in real-world problems,
laying the foundation for future academic or professional pursuits.

2.2 Problem Definition

Given a dataset of paired noisy and clean images, the task is to train a denoising autoencoder
model. The model should learn to reconstruct clean images from their noisy counterparts. The
architecture of the denoising autoencoder consists of an encoder component, which extracts
meaningful features from noisy input images, and a decoder component, which reconstructs
the clean images from these features. During training, the autoencoder is optimized to
minimize the reconstruction error between the output and the clean images.

2.3 Aim
The goal of this project on image denoising using autoencoder is to develop and evaluate an
effective denoising model capable of removing noise from images while preserving important
visual features
2.4 Objectives

• The main goal is to minimize the noise in the photos, which can be
brought on by transmission interference, environmental conditions, or
sensor constraints.

• Preservation of Image Quality: It's imperative to keep the image's key

elements and details when eliminating noise. Images that have been
denoised ought to hold onto their clarity, sharpness, and structural
integrity.

• Generalization: The model should be adaptable enough to work with a

variety of image datasets and novel or unique noise forms. It shouldn't
be unduly restricted to the noise patterns or training data used.

• Efficiency: For real-time applications or processing massive amounts of

photos, the denoising process should be computationally efficient.

• Robustness: The model must be able to withstand changes in image

quality, lighting.
 Chapter 3
Proposed System
Flowchart :

Figure 3.1 Flowchart

3.1 Working

1. Data Preparation:

Dataset: Collect or create a dataset containing pairs of noisy and clean images.
Preprocessing: Normalize the pixel values of the images and add noise to create the noisy
versions. The noise can be Gaussian noise, salt-and-pepper noise, or any other type of noise
depending on the application.

2. Model Architecture Design:

Encoder: The encoder component of the autoencoder consists of convolutional layers followed
by pooling layers to extract features from the input noisy images. The number of layers and
filter sizes can vary based on the complexity of the images and the noise level.
Decoder: The decoder component consists of convolutional layers followed by upsampling
layers to reconstruct the clean images from the learned features. The decoder mirrors the
architecture of the encoder.
Loss Function: The loss function used during training is typically Mean Squared Error (MSE)
or a perceptual loss function, which compares the reconstructed images with the clean images.

3. Training:

Data Split: Split the dataset into training, validation, and test sets. The training set contains
pairs of noisy and clean images used for training the autoencoder.
Training Process: Train the autoencoder using the training set. During training, the model
learns to minimize the difference between the noisy input images and their corresponding clean
versions.
Optimization: Utilize optimization techniques such as Stochastic Gradient Descent (SGD) or
Adam optimizer to update the model parameters and minimize the loss function.
Regularization: Apply regularization techniques like dropout or weight decay to prevent
overfitting and improve generalization.

4. Evaluation:

Quantitative Evaluation: Evaluate the trained autoencoder on the validation set using
quantitative metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity
Index (SSIM).
Qualitative Evaluation: Visualize the denoising performance by comparing the noisy input
images with the reconstructed clean images.

5. Inference:

Testing: Once trained, the denoising autoencoder can be used to remove noise from new,
unseen images. Feed the noisy images into the trained autoencoder, and obtain the denoised
output images.
3.2 Details of Hardware & Software
Hardware Requirements:

CPU: A multi-core processor (e.g., Intel Core i5 or higher) to handle the computational load
of training and evaluating machine learning models.

GPU (optional): A dedicated graphics processing unit (GPU) with CUDA support (NVIDIA
GeForce GTX or Quadro series) can significantly accelerate deep learning tasks.

RAM: At least 8 GB of RAM is recommended for efficient data processing and model training.

Storage: Sufficient storage space for storing datasets, model checkpoints, and software
dependencies.

Software Requirements:

Python: The primary programming language for implementing the digit recognition system
and its components. Version 3.x is recommended.

PyCharm: PyCharm is a Integrated Development Environment (IDE) specifically designed

for Python programming. It is tools for Python developers to write, test, debug, and deploy
their code efficiently.

Machine Learning Libraries:

NumPy: A fundamental package for numerical computing in Python, used for array
manipulation and mathematical operations.

Matplotlib: Visualization libraries for creating plots, charts, and graphs to analyse data and
evaluate model performance.

Operating System:
The system should be compatible with the chosen software stack. Common choices include
Windows.
3.3 Methodology
1. Data Preparation:

Clean Image Dataset: Acquire a dataset of clean images for training. Popular choices include
MNIST (handwritten digits), or your own collection of images.
Preprocessing: Preprocess the images to a format suitable for your deep learning framework.
This might involve resizing, normalization (scaling pixel values).

2. Noise Injection:

Noise Model: Choose a noise model to artificially corrupt the clean images. Common choices
include:

3. Building the Autoencoder:

Encoder: This network shrinks the image dimension (e.g., using convolutional and pooling
layers). You can experiment with the number of layers and filter sizes.
Latent Space: This is the bottleneck layer with a lower dimension compared to the input
image. It captures the essential features of the image.
Decoder: This network expands the latent representation back to the original image
dimension (e.g., using upsampling and convolutional layers).

4. Training the Model:

Loss Function: Choose a loss function to measure the reconstruction error between the noisy
input and the reconstructed clean image. Mean Squared Error (MSE) is a common choice.
Select an optimizer like Adam or RMSprop to update the model weights during training.
Training Process: Split your prepared data (noisy and clean image pairs) into training and
testing sets. Train the CAE by feeding noisy images as input and minimizing the
reconstruction error between the output and the corresponding clean images.

5. Evaluation:

Test Set: Evaluate the performance of your trained CAE on the unseen test set of noisy
images.
Metrics: Use quantitative metrics like Peak Signal-to-Noise Ratio (PSNR) or Structural
Similarity Index (SSIM) to compare the denoised images with the original clean images.
Visualization: Visualize some noisy images, their corresponding denoised outputs, and the
original clean images for qualitative assessment.

6. Enhancements:

Architecture Exploration: Experiment with different CAE architectures (e.g., deeper

networks, different activation functions).
Noise Type Exploration: Try training the model with different noise types (e.g., salt-and-
pepper) and compare the performance.
 Chapter 4
Code of Project
Code:
 Chapter 5
Result
Input:

Output:
 Chapter 6
Conclusion And Future Scope
In this mini-project demonstrates the effectiveness of autoencoder neural networks for image
denoising tasks. By training on pairs of noisy and clean images, the autoencoder learns to
reconstruct clean versions of the input images while removing noise. Through a
comprehensive methodology involving data preprocessing, model architecture design,
training, and evaluation, we successfully developed an image denoising system.

The results obtained from the trained autoencoder exhibit promising performance in removing
noise from images while preserving important features. Quantitative metrics such as PSNR and
SSIM provide objective measures of the model's denoising capability, while qualitative
evaluations visually demonstrate the improvement in image quality.

Moving forward, further optimization and exploration of autoencoder-based image denoising

techniques hold significant potential for real-world applications. Experimentation with
different architectures, loss functions, and training strategies could lead to even better
denoising performance. Additionally, applying these techniques to diverse datasets and
scenarios can validate the robustness and generalization capabilities of the developed model.
 Chapter 7
References
• Vincent, P., Larochelle, H., Bengio, Y., & Manzagol, P. A. (2008). Extracting and
composing robust features with denoising autoencoders. Proceedings of the 25th
International Conference on Machine Learning (ICML).

• Zhang, K., Zuo, W., Chen, Y., Meng, D., & Zhang, L. (2017). Beyond a Gaussian
denoiser: Residual learning of deep CNN for image denoising. IEEE Transactions on
Image Processing, 26(7), 3142-3155.

• Mao, X., Shen, C., & Yang, Y. B. (2016). Image restoration using convolutional auto-
encoders with symmetric skip connections. Advances in Neural Information Processing
Systems (NIPS), 1-9.

• Chen, Y., & Pock, T. (2017). Trainable nonlinear reaction diffusion: A flexible
framework for fast and effective image restoration. IEEE Transactions on Pattern
Analysis and Machine Intelligence, 39(6), 1256-1272.

• Lehtinen, J., Munkberg, J., Hasselgren, J., Laine, S., Karras, T., & Aila, T. (2018).
Noise2Noise: Learning Image Restoration without Clean Data. Proceedings of the 35th
International Conference on Machine Learning (ICML).

• Jain, V., & Seung, H. S. (2008). Natural image denoising with convolutional networks.
Advances in Neural Information Processing Systems (NIPS), 769-776.

• Zhang, Y., & Patel, V. M. (2017). Beyond a Gaussian denoiser: Residual learning of
deep CNN for image denoising. IEEE Transactions on Image Processing, 26(7), 3142-
3155.

• Yoo, D., Choi, M., Kim, J., & Yun, S. (2019). Deep learning-based image super-
resolution and denoising: A review. Journal of Imaging, 5(5), 52.

• Wang, Z., Liu, D., Yang, J., Han, W., & Huang, T. (2020). Learning to Denoise Images
with Deep CNNs. IEEE Transactions on Image Processing, 29, 4605-4616.

• Chen, Y., Zhong, Z., Xu, J., & Tao, D. (2020). Deep Learning in Image Denoising:
From Models to Algorithms. IEEE Transactions on Image Processing, 29, 4449-446