Parallel-Image-Processing

A C project that applies image-processing filters sequentially and in parallel using OpenMP. It currently implements grayscale conversion and binary thresholding, with benchmarks across different thread counts.

Features

• Load and save images using stb_image and stb_image_write.
• Apply grayscale conversion sequentially and in parallel.
• Apply binary thresholding sequentially and in parallel.
• Benchmark thread counts from 1 to 8.
• Measure execution time, speedup, and parallel efficiency.
• Verify that sequential and parallel outputs are identical.
• Save benchmark results in JSONL format.
• Generate performance charts using Python.

Project Structure

.
├── images/
│   ├── output/              # Processed images
│   └── image1.jpg           # Input image
├── include/                 # Header files and stb libraries
├── results/
│   ├── charts/              # Generated benchmark charts
│   └── benchmark_results.jsonl
├── scripts/
│   └── plot_results.py      # Benchmark visualization
├── src/                     # C source files
├── LICENSE
├── README.md
├── requirements.txt
└── tree.txt

Generated using Tree Printer.

Compile

Run these commands from the project root:

New-Item -ItemType Directory -Force build

gcc -std=c11 -Wall -Wextra -O2 -fopenmp -Iinclude `
    src/main.c src/image.c src/filters.c src/benchmark.c src/stb_impl.c `
    -o build/image_processor.exe

Run

The program accepts an input image, output path, filter name, and an optional threshold value.

Grayscale conversion:

.\build\image_processor.exe images/image1.jpg images/output/grayscale.png grayscale

Binary thresholding (default threshold: 128):

.\build\image_processor.exe images/image1.jpg images/output/threshold.png threshold

Binary thresholding with a custom threshold:

.\build\image_processor.exe images/image1.jpg images/output/threshold_180.png threshold 180

The threshold value must be an integer between 0 and 255.

Otsu binarization:

.\build\image_processor.exe images/image1.jpg images/output/threshold_180.png otsu

All filters:

.\build\image_processor.exe images/image1.jpg images/output/result.png all

For each filter, the program benchmarks the sequential implementation and then the parallel implementation using 1 to 8 threads. It checks the outputs and appends the measurements to:

results/benchmark_results.jsonl

Generate Charts

Install the required Python packages:

pip install -r requirements.txt

Generate the benchmark charts:

python scripts/plot_results.py

The charts are saved in results/charts/, with separate charts for each filter.

Results Analysis

For a detailed discussion of benchmark results, performance scaling, and the observed speedup behavior, see Results Analysis.

License

See LICENSE for license information.