SAR-Image-Processing

This project implements a comprehensive pipeline for Synthetic Aperture Radar (SAR) image processing, focusing on performance comparisons between sequential (CPU), OpenMP (multi-threaded), and CUDA (GPU-accelerated) implementations. It includes detailed profiling, execution time analysis, and visualization tools to assess the efficiency of each approach.

SAR-Image-Processing/
├── images/
│   └── sample1.jpg
├── main.cpp
├── openmp_implementation.cpp
├── pixel_imp.cpp
├── sample_testing.cpp
├── compare_execution_times.py
├── img_size_vs_exec_time.py
├── plot_graph.py
├── cuda_execution_times.csv
├── openmp_times.csv
├── pixel_implementation.csv
├── image_size_vs_time.csv
├── execution_time_comparison.png
├── img_size_vs_seq_exec_time.png
├── gprof_report.txt
├── gprof.png
├── LICENSE
└── README.md

🚀 Features

• Image Processing Pipeline: Includes grayscale conversion, rotation, flipping, HSV conversion, brightness adjustment, clipping, histogram equalization, Wiener filtering, and Gaussian filtering.

• Multiple Implementations:

  • Sequential (CPU): Baseline OpenCV implementation in main.cpp.
  • Sequential (CPU): Baseline Manual pixel implementation in pixel_imp.cpp.
  • OpenMP (Multi-threaded CPU): Parallelized version in openmp_implementation.cpp.
  • CUDA (GPU-accelerated): High-performance version in https://www.kaggle.com/code/venkatasagarmaddela/hpc-project.

• Performance Analysis:

  • Execution time measurements for each function.
  • Profiling reports using gprof.
  • Visualization scripts to compare performance across implementations.

🛠️ Setup & Installation

1. Clone the Repository:

   git clone https://github.com/SagarMaddela/SAR-Image-Processing.git
   cd SAR-Image-Processing

2. Install Dependencies:

   • Ensure you have OpenCV installed for image processing.
   • For CUDA implementation, ensure CUDA toolkit is installed or Use Kaggle Notebook or Use Google Colab.
   • Python 3 with matplotlib and pandas for plotting scripts.

3. Build the Project:

   • Sequential OpenCV Version:

     g++ main.cpp -o main `pkg-config --cflags --libs opencv4`

• Sequential Manual Pixel Version:

  g++ pixel_imp.cpp -o pixel_imp `pkg-config --cflags --libs opencv4`

• OpenMP Version:

  g++ -fopenmp openmp_implementation.cpp -o openmp `pkg-config --cflags --libs opencv4`

• CUDA Version:

  !nvcc final_cuda_implementation -o cuda `pkg-config --cflags --libs opencv4`

🧪 Usage

1. Run Sequential Implementation:

   ./main

2. Run OpenMP Implementation:

   ./openmp

3. Run CUDA Implementation:

   ./cuda

4. Generate Performance Plots:

   python compare_execution_times.py
   python img_size_vs_exec_time.py
   python plot_graph.py

📊 Results & Analysis

• Execution Time Comparison:

  • execution_time_comparison.png: Visual comparison of execution times across implementations.
  • img_size_vs_seq_exec_time.png: Execution time vs. image size for sequential implementation.

• Profiling Reports:

  • gprof_report.txt: Detailed profiling report generated using gprof.
  • gprof.png: Visual representation of profiling data.

• CSV Data:

  • cuda_execution_times.csv, openmp_times.csv, pixel_implementation.csv, image_size_vs_time.csv: Raw execution time data for further analysis.