houghvg 0.2.0

A Hough Transform Toolbox Based on Virtual Grids

HoughVG : Hough transform on Virtual Grids: Hough transform toolbox for straight-lines detection and fingerprints recognition

HoughVG is a Virtual Grid-Based Hough transform toolbox for straight-lines detection and fingerprints recognition. It brings together several innovative variants of the Virtual Grid-Based Hough transform, including the rectangular, triangular, hexagonal and octagonal Hough transforms for straight-line detection, as well as the generalized Hough transform using virtual rectangular grid, specially adapted for fingerprint recognition.

This version of HoughVG is implemented in C++ and exposed to Python using nanobind.

At this stage, fingerprint recognition and parallel processing are not yet available in this release. These features will be added in a future version.

Installation From PyPI

Install the package with pip:

python -m pip install houghvg

Usage

Straight-lines detection

import time
import cv2
import houghvg
import numpy as np

PI = np.pi
IMAGE_PATH =  "univ.PNG"
THRESHOLD = 31

def draw_hough_lines(image, lines, color):
    output = image.copy()
    if lines is None:
        return output

    for line in lines:
        rho = float(line[0])
        theta = float(line[1])
        cos_theta = np.cos(theta)
        sin_theta = np.sin(theta)

        pt1 = (
            int(round(rho * cos_theta + 1000 * (-sin_theta))),
            int(round(rho * sin_theta + 1000 * cos_theta)),
        )
        pt2 = (
            int(round(rho * cos_theta - 1000 * (-sin_theta))),
            int(round(rho * sin_theta - 1000 * cos_theta)),
        )
        cv2.line(output, pt1, pt2, color, 2, cv2.LINE_AA)

    return output

def main():
    image = cv2.imread(str(IMAGE_PATH), cv2.IMREAD_COLOR)
    if image is None:
        raise RuntimeError(f"Impossible de charger {IMAGE_PATH}")

    gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    edges = cv2.Canny(gray, 100, 200)

    t0 = time.perf_counter()
    houghvg_lines = houghvg.HoughLines(
        edges,
        grid_type=houghvg.GridType.OCTOGONAL,
        rho=1.0,
        theta=PI / 180.0,
        threshold=THRESHOLD,
        cell_width_base=4.0,
        cell_height=5.0,
        gamma=2.0,
        rate=0.4,
        use_parallel=False,
        return_numpy=True,
    )
    
    img_houghvg_lines = draw_hough_lines(image, houghvg_lines[:, :2], (0, 255, 0))
    
    cv2.imshow("HoughVG : Detected lines", img_houghvg_lines)
    print("Appuie sur une touche dans une fenetre OpenCV pour fermer.")
    cv2.waitKey(0)
    cv2.destroyAllWindows()

Features

• Rectangular, triangular, hexagonal, and octagonal virtual grids.
• Python bindings built with nanobind.
• Optional NumPy return mode: return_numpy=True returns an Nx3 float32 array containing [rho, theta, votes].
• Example script comparing houghvg.HoughLines and cv2.HoughLines visually and temporally.

Requirements

This package contains a native C++ extension. The current Windows build links against OpenCV 4.9