rvmn 0.1.0

RVMN directional edge detection filters and comparison utilities.

rvmn

RVMN is a small Python package for directional edge detection using custom RVMN-style masks, plus Sobel/Prewitt comparison helpers.

Install

pip install rvmn

For local development from this folder:

pip install -e ".[dev,plots]"

Python Usage

import cv2
import rvmn

image = cv2.imread("image.png")
gray = rvmn.to_grayscale(image)

result = rvmn.apply_rvmn(gray, lam=0.1, rho=0.9, theta=0.3)
sobel = rvmn.apply_sobel(gray)
prewitt = rvmn.apply_prewitt(gray)

metrics = rvmn.compute_metrics(gray, result)
print(metrics.psnr, metrics.rmse)

Search for the best parameter combination:

search = rvmn.search_best_params(gray)

print(search.params)
print(search.metrics.psnr)
cv2.imwrite("rvmn_output.png", search.image)

Compare RVMN, Sobel, and Prewitt in one call:

comparison = rvmn.compare_edges(gray)

print(comparison.rvmn.params)
print(comparison.sobel_metrics.psnr)
print(comparison.prewitt_metrics.psnr)

Command Line

rvmn image.png --output results

This writes rvmn_output.png, sobel_output.png, and prewitt_output.png.

Use fixed RVMN parameters instead of a full search:

rvmn image.png --lambda 0.1 --rho 0.9 --theta 0.3 --output results

Print machine-readable metrics:

rvmn image.png --json

Build

python -m build

The build configuration includes only src/, tests/, README.md, and pyproject.toml in the source archive, and only the rvmn package in the wheel. Generated notebooks, images, PDFs, result folders, caches, and old dist/ contents are excluded.

Notes

The coefficient formulas are packaged from the current notebook implementation. Keep citation and formula details close to your paper/report when publishing.