Image Filtering and Performance Comparison Notebook

This project implements three common image filters and compares their performance using three different approaches:

• Gaussian Blur: Smooths the image using a Gaussian kernel.
• Sobel Edge Detection: Detects edges by computing gradients in the image.
• Median Filter: Reduces noise by replacing each pixel with the median value in its neighborhood.

These filters are implemented in:

1. Pure Python – Using nested lists and loops.
2. NumPy – Leveraging optimized vectorized operations and SciPy functions.
3. Cython – Combining Python syntax with C-level performance optimizations.

The notebook measures execution times and estimates MFLOPS (Millions of Floating-Point Operations per Second) for each implementation, displaying interactive plots and console summaries.

---

Table of Contents

• Project Overview
• Implemented Filters
• Implementations
• Usage Instructions
• Cython Compilation
• Viewing Results
• License

---

Project Overview

This notebook is designed to demonstrate the differences in performance between various implementations of image filters. The focus is on:

• Comparing pure Python, NumPy, and Cython approaches.
• Analyzing performance metrics such as execution time and MFLOPS.
• Visualizing processed images and performance data.

---

Implemented Filters

1. Gaussian Blur

   • Applies a Gaussian kernel to smooth the image.
   • Reduces noise and detail.

2. Sobel Edge Detection

   • Detects edges by computing gradients in the horizontal and vertical directions.
   • Combines gradients to produce the edge magnitude.

3. Median Filter

   • Reduces noise by replacing each pixel with the median value of its surrounding pixels.
   • Particularly effective for removing salt-and-pepper noise.

---

Implementations

Pure Python

• Uses nested loops and lists.
• Manually handles padding and kernel application.
• Simple to understand but slower due to Python loop overhead.

NumPy

• Utilizes vectorized operations on NumPy arrays.
• Employs SciPy functions (such as convolve and median_filter) for optimized processing.
• Significantly faster than pure Python due to low-level optimizations.

Cython

• Combines Python code with static type declarations.
• Compiles to C for improved performance.
• Offers a substantial speedup over pure Python, though in these tests, NumPy’s optimized routines sometimes outperform the Cython version.

---

Usage Instructions

1. Install Dependencies
   Ensure you have the required libraries installed:

   pip install pillow matplotlib numpy scipy plotly cython setuptools