Image Denoise Filter Tool with Streamlit

A simple image processing web app built using Streamlit application designed to demonstrate various image processing techniques, including Gaussian Blur, Median Filtering, PSNR (Peak Signal-to-Noise Ratio), and SSIM (Structural Similarity Index). This app allows users to upload an image and apply these filters, viewing the processed images alongside their original counterpart. It also provides metrics to quantitatively compare the differences between the original and filtered images. As well as, view histograms of RGB color channels for each image.

---

Table of Contents

• Image Denoise Filter Tool with Streamlit
  • Table of Contents
  • Project Overview
  • Features
  • Installation
  • Usage
  • Example Screenshot and Video
  • Filtering Techniques and Metrics Definitions
  • Example Use Case
  • Summary
  • MIT License

---

Project Overview

This project is a web-based image processing application built using Streamlit, allowing users to:

• Upload images in .jpg, .jpeg, or .png format.
• Apply Gaussian Blur and Median Filtering techniques.
• View processed images side-by-side with the original.
• Calculate metrics like PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index).
• Save the processed images to disk.

The application uses OpenCV for image processing and Streamlit for the web interface. Images are loaded as NumPy arrays, processed using OpenCV filters (Gaussian Blur & Median Filter), and displayed in RGB format. PSNR and SSIM metrics are computed to evaluate the quality of the processed images. Histograms are visualized using Matplotlib.

---

Features

Feature	Description
✅ Image Upload	Supports JPG, JPEG, and PNG formats.
✅ Gaussian Blur	Apply blur with adjustable kernel size and sigma.
✅ Median Filter	Apply noise reduction using median filtering.
✅ PSNR & SSIM Metrics	Calculate image quality metrics for comparison.
✅ Histogram Visualization	View RGB histograms of original, blurred, and filtered images.
✅ Save Processed Images	Export processed results to disk.

---

Installation

To run this application locally, follow these steps:

1. Install dependencies using pip:

   pip install streamlit numpy opencv-python pillow scikit-image matplotlib

2. Save the code in a file, for example: app.py.

3. Run the app with Streamlit:

   streamlit run app.py

---

Usage

1. Upload an image (JPG/JPEG/PNG): Use the file uploader to load an image in JPG, JPEG, or PNG format.
2. Adjust Parameters for Image Processing:
   • Adjust Gaussian Blur kernel size and sigma
   • Adjust Median Filter kernel size
3. View Processed Results: After adjusting, you'll see three side-by-side images:
   • Original
   • Gaussian Blurred Image
   • Median Filtered Image
4. Compute Metrics: The app will automatically calculate and display the following metrics for each processed image:
   • PSNR (Peak Signal-to-Noise Ratio) in dB
   • SSIM (Structural Similarity Index)
5. Save Processed Images:You can click "Save Processed Images" to save both results as JPEG files.
6. View Color Histograms: There's an option to view histograms of RGB color channels for each image. Display histograms of the RGB channels for the original, Gaussian-blurred, and median-filtered images.

---

Example Screenshot and Video

denoise-filter-app.mp4

---

Filtering Techniques and Metrics Definitions

• Gaussian Blur
  A smoothing filter that reduces noise by averaging pixel values with a Gaussian distribution kernel. Ideal for reducing fine details but can blur edges.

• Median Filter
  A non-linear filter that replaces each pixel with the median of its neighbors, effective against salt-and-pepper noise while preserving edges better than Gaussian filters.

Peak Signal-to-Noise Ratio

• PSNR is calculated using the Mean Squared Error (MSE) between the original and processed images.

• The formula:

  $\text{PSNR} = 10 \cdot \log_{10}\left( \frac{\text{MAX\_PIXEL}^2}{\text{MSE}} \right)$

  • MAX_PIXEL: Maximum possible pixel value (e.g., 255 for 8-bit images).
  • MSE: Mean Squared Error between the original and reconstructed image.
  • log₁₀: Logarithm base 10.

What it tells you:

• Higher PSNR = Better Quality: Indicates minimal noise or distortion.

• Lower PSNR = More Distortion: Suggests significant degradation in image quality (e.g., due to compression, blurring, or filtering).

• Use Case: Often used for objective evaluation of image compression algorithms or noise reduction techniques.

Structural Similarity Index (SSIM)

• SSIM evaluates how similar two images are in terms of their structure, luminance (brightness), and contrast, mimicking human visual perception. It is more aligned with subjective image quality assessment than PSNR.

How it works:

• SSIM compares the local structure of patches in both images using a windowed approach.

• The formula:
  $\text{SSIM}(x, y) = \frac{(2\mu_x\mu_y + C_1)(2\sigma_{xy} + C_2)}{(\mu_x^2 + \mu_y^2 + C_1)(\sigma_x^2 + \sigma_y^2 + C_2)}$

  • where μ is the mean,
  • σ is the standard deviation,
  • and C₁, C₂ are constants to stabilize values.

What it tells you:

• Higher SSIM = Better Structural Similarity: Indicates that the processed image retains key structural details of the original (e.g., edges, textures).
• Lower SSIM = Structural Differences: Suggests significant changes in visual structure (e.g., blurring, edge loss, or artifacts).
• Use Case: Ideal for assessing perceptual quality in applications like image restoration, compression, and filtering.

Key Differences between PSNR and SSIM

Metric	Focus	What It Tells You
PSNR	Pixel-level differences	Noise or distortion in the image (objective).
SSIM	Structural similarity	How similar the images look to a human viewer.

---

Example Use Case

• If you apply Gaussian Blur to an image:
  • PSNR will drop significantly (high noise/blur).
  • SSIM may decrease more gradually, reflecting how much structure is preserved.
• If you apply Median Filtering (to reduce noise):
  • PSNR improves if noise is reduced.
  • SSIM might remain stable or slightly improve, as median filtering preserves edges better than Gaussian blur.

---

Summary

• Use PSNR for quick, objective assessments of pixel-level quality.
• Use SSIM to evaluate how well the image retains its visual structure and perceptual fidelity.
• Together, these metrics provide a comprehensive view of image degradation or enhancement during processing.

---

MIT License

This project is licensed under the MIT License.

---