Python Image Morpher (PIM) is a program that can take two images and blend them to whatever extent or precision that you like! It is designed to emulate some of Python's OpenCV image processing from scratch without reference.

This project began in Spring 2019 (detailed in readme_general.pdf, readme_phase_1.pdf, and readme_phase_2.pdf), requiring the usage of delaunay triangulation, projective/affine transformation, application of projections through matrices, masking, and alpha blending - all of which are still maintained - in order to accomplish image morphing. Thus, this repository is more of a 'proof of concept' than it is the most efficient way of accomplishing the given task of image morphing.

So far, this program has only been tested on separate Windows environments; if it does not already, later releases are likely to support Mac and Linux.

Installation:

This program has dependencies that do not come packaged with Python 3. To install the required modules, enter the following command using the terminal supplied by Windows or your choice of IDE (such as PyCharm):

pip install -r requirements.txt

Alternatively, for each module below that your machine does not have installed, enter the respective command(s):

PyQt5 - pip install pyqt5

SciPy - pip install scipy

NumPy - pip install numpy

Imageio - pip install imageio

Matplotlib - pip install matplotlib

If pip, for whatever reason, is not installed on your machine, enter the following line in a terminal:

curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py

Then navigate to the folder where get-pip.py was downloaded and enter the following line in a terminal:

python get-pip.py

Usage:

• Run MorphingApp.py either through the terminal or using an IDE
• Use the graphical interface to select your two images to morph
• Click on points of interest in your images to create correspondences
• When satisfied, click on blend and observe the result!

Proof of Concept - Perspective Shifting

Development 'To-Do' List:

• Change: Improved Drawing Performance
  • There is currently a small delay with point placement when working with larger images.
• Feature: Automatic Correspondence Determination
  • The user may click a button to have PIM automatically generate points by scanning for similar features between images
• Feature: Resizable Image Points
  • Placed points now scale in size with their respective image (i.e. smaller images will use smaller points)
  • Additionally, the user may manually scale point size with a slider on the GUI

If you encounter an error, a bug, or if you simply wish to request a change/feature, please file an issue using the tracker that GitHub provides, here.

If you like what you see, feel free to contact me on LinkedIn.