morecvutils 0.9.2

Computer Vision utilities, Cohen-Sutherland line clipping, OpenCV plot helpers for Optical Flow and Blob Analysis, AVI codec helpers

CVutils

Author:

Michael Hirsch

License:

MIT

Prereq:

OpenCV 2 or OpenCV 3

Contents

• CVutils

  • Install

    • Fortran Build (optional)

  • Usage

    • Python

    • Fortran

    • Julia

  • Functions

Misc. algorithms useful for computer vision.

Install

python setup.py develop

Fortran Build (optional)

If you want to use the Fortran Cohen-Sutherland line clipping modules directly (optional):

cd bin
cmake ..
make

Usage

The main difference with textbook implementations is that I return a sentinel value (NaN, None, nothing) if there’s no intersection of line with box.

Python

import morecvutils.lineclipping as lc

x3,y3,x4,y4 = lc.cohensutherland((xmin, ymax, xmax, ymin, x1, y1, x2, y2)

If no intersection, (None, None, None, None) is returned.

Fortran

lineclipping.f90 has two subroutines. Pick Ccohensutherland if you’re calling from C/C++/Python, which cannot tolerate assummed-shape arrays. It’s a slim wrapper to cohensutherland which is elemental (can handle scalar or any rank array).

Fortran programs will simply use

use lineclipping
call cohensutherland(xmin,ymax,xmax,ymin,x1,y1,x2,y2)

The arguments are:

INPUTS
------
xmin,ymax,xmax,ymin:  upper left and lower right corners of box (pixel coordinates)

INOUT
-----
x1,y1,x2,y2:
in - endpoints of line
out - intersection points with box. If no intersection, all NaN

Julia

Simliar to Python, except nothing is returned if no intersection found.

cohensutherland(xmin, ymax, xmax, ymin, x1, y1, x2, y2)

Functions

function	description
lineClipping.jl	Cohen-Sutherland line clipping algorithm for Julia. Input scalars, output intersection length, or None if no intersection.
lineclipping.f90	Cohen-Sutherland line clipping algorithm for Fortran. Input scalars or arrays, output intersections.
lineClipping.py	Cohen-Sutherland line clipping algorithm for Python. Input scalars, output intersection length, or None if no intersection.
draw_flow()	given a 2-D complex Numpy array of optical flow flow, draw flow vectors with arrows
draw_hsv()	make a colored HSV image corresponding to flow direction and intensity at each point
connectedComponents.py	given a binary image morphed and the blobdet from setupblob(), along with img, do connected components analysis
OpticalFlow_Matlab_vs_Python.py	using Horn-Schunck optical flow estimation with OpenCV in Python. Not so obvious from the docs, and with notes on how to make this match Matlab’s vision.opticalFlowHS method. Install Matlab Engine for Python