Overview
Image convolutions like Gaussian blur, Sobel edge detection, and sharpening are embarrassingly parallel across pixels, but boundary rows need data from their neighbours. This project uses MPI to split a large image into row strips, distribute them across processes, exchange halo rows between adjacent ranks, run the convolution locally, then gather the result back to rank 0 for output.
Highlights
- MPI scatter/gather with configurable halo rows for boundary handling
- Applied Gaussian blur, Sobel edge detection, and sharpen kernels
- Validated correctness across multiple image sizes and process counts
- Kernels loaded from file so the same binary supports arbitrary filter sizes
Tech Stack
C / C++
MPI
OpenCV
Linux