pocpy (Phase Only Correlation in Python)

This repository contains implementations of phase only correlation and rotation invariant phase only correlation in python.

Installing pocpy

$ pip install pocpy

Usage

import cv2
import numpy as np

import pocpy

f = cv2.imread("reference.png", cv2.IMREAD_GRAYSCALE).astype(np.float64)
g = cv2.imread("target.png", cv2.IMREAD_GRAYSCALE).astype(np.float64)

# translation only
peak, dy, dx = pocpy.poc(f, g)

# translation, rotation and scale
peak, dy, dx, angle, scale = pocpy.ripoc(f, g)

• f and g are 2D arrays of the same shape.
• poc returns the sub-pixel translation (dy, dx) of g relative to f.
• ripoc assumes g is f translated by (dy, dx) and then rotated by angle degrees (counter-clockwise, as in cv2.getRotationMatrix2D) and scaled by scale about the image center.
• peak is the height of the correlation peak and can be used as a similarity measure.
• Rotation is estimated within ±90 degrees, because the amplitude spectrum used by ripoc cannot distinguish angle from angle ± 180.

Development

$ uv sync
$ uv run pytest
$ uv run ruff check .
$ uv run ruff format --check .