CudaCanny Edge Detector

A high-performance implementation of the Canny edge detector with both CPU and CUDA GPU paths. It loads PGM images, runs Gaussian smoothing, gradient, non-maximum suppression, and hysteresis, then saves CPU and GPU edge maps while reporting timing and speedup.

Features

• CPU and CUDA implementations side-by-side
• Benchmarking with timing and speedup
• Simple Makefile build

Requirements

• CUDA Toolkit (nvcc)
• GCC/Clang for C sources

Build

make

Binaries are placed in bin/.

Run

./bin/canny canny/pics/pic_large.pgm 2.5 0.25 0.5

Adjust sigma, tlow, thigh as needed. Example images are in canny/pics/.

Project Layout

• src/: C/CUDA sources
• include/: headers
• bin/: compiled binary
• build/: object files
• canny/pics/: sample images (PGM)