Preprocessing video

Overview

Teaching: 5 min
Exercises: 10 min

Questions

• How is it possible to improve the video before analysing?

Objectives

• Learn to trim, crop and downsample a video.

• Learn some of the specialized functions in MGT.

Working with structures

For the more advanced examples we will change from working directly on video files to handling a matlab structure containing a pointer to the file.

mg = mgread('dance.avi');

From here you can refer to mg instead of the name of the video file.

Trimming the files

If we only want to analyse a part of this recording, we may choose to extract a segment of 20 seconds from 10 seconds to 30 seconds of the original video. Here MGT will handle the trimming of all related files and data:

mgseg = mgreadsegment(mg,10,30);

Here mgseg is a new (and complete) structure that can be used for further processing.

Cropping video

To further save some computational time, and focus more particularly on a particular part of the video, we may also look only at the region of interest of the video. The mgvideocrop function will crop specified region on each frame. Start by using:

mgcrop = mgvideocrop(mg);

The user can “draw” the region on the video frame followed by right-clicking to start the export.

Rotating

Sometimes you may want to rotate the video before doing an analysis. For example, analysing standing people may be easier to do with a video recorded in portrait mode. Then it may be useful to rotate the video in Matlab. This can be done like this:

mgrot = mgvideorotate(mg,90);

Downsampling

Analysing large, full-size video files can be very time-consuming. Therefore it may sometimes be useful to downsample the video file before moving on. This can be done like this:

mgsam = mgvideosample(mg,[2,2],'samplevideo.avi');

This reduces the framesize of the video with a factor of 2 in both rows and columns. It also writes the downsampled video to a new file ‘samplevideo.avi’.

Run your analysis

After doing all of the above, you may want to run your analysis on the output, say:

mgan = mgmotion(mgsam);

Colour analysis

The plan is to implement a colour option in all functions. For now, however, you can enable colour processing by setting this flag:

mg.video.mode.color = 'On';

Then you can run

mgmotion(mg,'Diff','Regular',0.1);

to create a motion video and motiongrams in colour.

Colour inversion

Sometimes you may want to invert colours in the output files, particularly if you want to print something on paper. Then you can set this flag:

mg.video.mode.convert = 'On'

and possibly also revert back to using greyscales:

mg.video.mode.color = 'Off';

before running the analysis:

mgmotion(mg,'Diff','Regular',0.1);

and you will get black on white, instead of white on black.

Key Points

• A video is trimmed in time, cropped in space, and downsampled in pixels.

• There are a number of things to set and modify in the code, most of which are documented in the help files.