movie-crop

Crop movies using Python.

Use your mouse to select a region of a frame in the movie, and then you can save the cropped movie or a sequence of images, depending on which parameters you choose. Uses OpenCV and numpy.

Usage

In movie_crop.py, in the first two sections set up imports and set the parameters (see code for details).
Run through remaining sections of the script:
- Define functions
- Open movie and view metadata (optional)
- Preview movie (optional)
- Open frame to select cropping parameters
- Save cropped data (either image stack or as another movie)

Notes

OpenCV does not play nicely with Jupyter so I recommend not using this with Jupyter.
To save an image stack, set the save_filetype parameter to png. To save as a movie, set it to avi or mp4, and also set the compression codec. There are already values set, and suggestions provided, in the parameters section of the script.
When you save without loss ('FFV1' codec) the files can get huge. For smaller files, I recommend the 'DIVX' codec which compresses really well but retains high quality.
This is a simple tool: if you want tons of choices/features, I recommend the command-line tool ffmpeg. movie-crop is basically a front-end wrapper for that. Also, ImageJ is really good too, but often there are quite a few codecs it will not work with, and I do not know how to control it.
Clip is from the open-source cartoon Big Buck Bunny (https://en.wikipedia.org/wiki/Big_Buck_Bunny).
The standard command-line tool for movie editing is ffmpeg, and opencv typically uses it in the back end. To check in your case you can use print(cv2.getBuildInformation()) and look for Video I/O.
When making this I did a bunch of experiments saving 500 frames to movies to see which codecs went best with different file types (focusing on mp4 and avi because they are the most common). Results, and other general notes, are in supplemental.md.

About

Part of an analysis pipeline developed with the Neurobehavioral Core at NIEHS: https://www.niehs.nih.gov/research/atniehs/facilities/neurobehavioral/index.cfm

To do

Incorporate pims(http://soft-matter.github.io/pims/v0.5/)
Set up to work with tiff stacks
Make batch version that applies to all movies in a folder.
Add a rescale option.
Option to read/save greyscale (imread(image_path, cv2.IMREAD_GRAYSCALE))
Clicking X should work because this is 2020. But opencv is weird: https://stackoverflow.com/questions/35003476/opencv-python-how-to-detect-if-a-window-is-closed Maybe just use matplotlib for display?
Change quality of encoding? https://stackoverflow.com/questions/61824667/does-videowriter-prop-quality-do-anything-in-opencv

To think about doing