Visual Servoing Platform  version 3.5.1 under development (2022-05-22)
Tutorial: How to manipulate a video or a sequence of images

Introduction

In this tutorial you will learn how to manipulate a video or a sequence of successives images in order to rename the images, convert images format, or select some images that will constitute a dataset typically for deep learning purpose.

Note that all the material (source code and images) described in this tutorial is part of ViSP source code and could be downloaded using the following command:

$ svn export https://github.com/lagadic/visp.git/trunk/tutorial/image

Prerequisites

Sequence of images acquisition reminder

In ViSP you will find Tutorial: Image frame grabbing that shows how to acquire a sequence of images.

  • For example, if you are running Ubuntu on a laptop, you can acquire a sequence running:
    $ cd $VISP_WS/visp-build/tutorial/grabber
    $ ./tutorial-grabber-v4l2 --seqname /tmp/myseq/png/I%04d.png
    Use device : 0
    Recording  : enabled
    Display    : enabled
    Record mode: continuous
    Record name: /tmp/myseq/png/I%04d.png
    Warning: cannot set input channel to 2
    Image size : 640 480
    
  • At this point you have to click to start saving the images
    Create directory "/tmp/myseq/png"
    Started sequence saving: /tmp/myseq/png/I%04d.png
    
  • A right click allows to quit
    Wait to finish saving images...
    Receive cancel during gray image saving.
    
  • Here we have all the images saved in /tmp/myseq/png folder
    $ ls -1 /tmp/myseq/png
    I0001.png
    I0002.png
    I0003.png
    I0004.png
    I0005.png
    I0006.png
    I0007.png
    I0008.png
    I0009.png
    I0010.png
    I0011.png
    I0012.png
    I0013.png
    I0014.png
    I0015.png
    I0016.png
    I0017.png
    I0018.png
    I0019.png
    I0020.png
    I0021.png
    ...
    

Video or image sequence manipulation

We provide a tool which source code is given in tutorial-image-manipulation.cpp that allows video manipulation. To see what could be done, just get the helper message:

$ cd $VISP_WS/visp-build/tutorial/image
$ ./tutorial-video-manipulation --help

Visualization

The tutorial tutorial-image-manipulation.cpp allows to visualize a video or a sequence of images.

  • To visualize an mpeg video part of ViSP data set, you may run:
    $ cd $VISP_WS/visp-build/tutorial/image
    $ ./tutorial-video-manipulation --in ${VISP_INPUT_IMAGE_PATH}/video/cube.mpeg
    
  • To visualize a sequence of successive images, like the one acquired in Sequence of images acquisition reminder section, you may rather run:
    $ cd $VISP_WS/visp-build/tutorial/image
    ./tutorial-video-manipulation --in /tmp/myseq/png/I%04d.png
    

Renaming and/or image format conversion

The tutorial tutorial-image-manipulation.cpp allows also to rename and convert the images format.

  • The following example shows how to convert all the images from png to jpeg and also rename the images:
    $ cd $VISP_WS/visp-build/tutorial/image
    $ ./tutorial-video-manipulation --in /tmp/myseq/png/I%04d.png --out /tmp/myseq/jpeg/image-%04d.jpeg
    Input video
      Video name     : /tmp/myseq/png/I%04d.png
      Video dimension: 640 480
      First image    : 1
      Last image     : 49
      Framerate (fps): 30
    Output video
      Video name     : /tmp/myseq/jpeg/image-%04d.jpeg
      First image    : 1
      Stride         : 1
      Y8 gray images : no (same as input)
    Other settings
      verbatim        : disabled
      Select frames  : disabled
    
  • Renamed and converted images are now available in /tmp/myseq/jpeg folder
    $ ls -1 /tmp/myseq/jpeg/
    image-0001.jpeg
    image-0002.jpeg
    image-0003.jpeg
    image-0004.jpeg
    image-0005.jpeg
    image-0006.jpeg
    image-0007.jpeg
    image-0008.jpeg
    image-0009.jpeg
    image-0010.jpeg
    image-0011.jpeg
    image-0012.jpeg
    image-0013.jpeg
    image-0014.jpeg
    image-0015.jpeg
    image-0016.jpeg
    image-0017.jpeg
    image-0018.jpeg
    image-0019.jpeg
    image-0020.jpeg
    image-0021.jpeg
    ...
    

There is also an extra option --out-first-frame <index> that allows to modify the index of the output video first frame

  • For example, to start output numbering at 100, you may run
    $ ./tutorial-video-manipulation --in /tmp/myseq/png/I%04d.png --out /tmp/myseq/jpeg/image-%04d.jpeg --out-first-frame 100
    
  • The index of the images becomes:
    $ ls -1 /tmp/myseq/jpeg
    image-0100.jpeg
    image-0101.jpeg
    image-0102.jpeg
    image-0103.jpeg
    image-0104.jpeg
    image-0105.jpeg
    image-0106.jpeg
    image-0107.jpeg
    image-0108.jpeg
    image-0109.jpeg
    image-0110.jpeg
    image-0111.jpeg
    image-0112.jpeg
    image-0113.jpeg
    image-0114.jpeg
    image-0115.jpeg
    image-0116.jpeg
    image-0117.jpeg
    image-0118.jpeg
    image-0119.jpeg
    image-0120.jpeg
    

Moreover, there is also an other extra option --out-gray that allows to save output images in Y8 gray level images.

  • For example, considering that /tmp/myseq/png/I%04d.png input images are color images, if you want to convert them in Y8 gray, you may run:
    $ ./tutorial-video-manipulation --in /tmp/myseq/png/I%04d.png --out /tmp/myseq/gray-jpeg/gray-image-%04d.jpg --out-gray
    
  • Finally, there is also the --out-stride option that allows to keep one image over n in the resulting output video. For example, if your input image sequence has 40 images and you want to create a new image sequence temporally subsampled with only 20 images, you can use this option like:
    $ ./tutorial-video-manipulation --in /tmp/myseq/png/I%04d.png --out /tmp/myseq/png-stride-2/I%04d.png --out-stride 2
    

Images extraction from a video

The tutorial tutorial-image-manipulation.cpp allows also to extract images from a video file.

  • For example to extract the images from an mpeg video part of ViSP data set and create a sequence of successive images, you may run:
    $ ./tutorial-video-manipulation --in ${VISP_INPUT_IMAGE_PATH}/video/cube.mpeg --out /tmp/cube/jpeg/image-%04d.jpeg
    
  • Images are available in /tmp/cube/jpeg folder:
    image-0001.jpeg
    image-0002.jpeg
    image-0003.jpeg
    image-0004.jpeg
    image-0005.jpeg
    image-0006.jpeg
    image-0007.jpeg
    image-0008.jpeg
    image-0009.jpeg
    image-0010.jpeg
    ...
    
  • You can then replay the image sequence using:
    ./tutorial-video-manipulation --in /tmp/cube/jpeg/image-%04d.jpeg
    

Images selection to create a new video or sequence

The tutorial tutorial-image-manipulation.cpp allows also to extract some images selected by the user during visualisation by user click. This feature could be useful to extract the images that will be part of a deep leaning data set.

  • To create a new video from selected images you may add --select option, like:
    $ ./tutorial-video-manipulation --in ${VISP_INPUT_IMAGE_PATH}/video/cube.mpeg --out /tmp/cube/jpeg/image-%04d.jpeg --select
    
  • Here if the user click four times in the video, you will get a new sequence with 4 successive images
    $ ls -1 /tmp/cube/jpeg 
    image-0001.jpeg
    image-0002.jpeg
    image-0003.jpeg
    image-0004.jpeg
    

Next tutorial

You are now ready to see how to continue with Tutorial: How to modify an image to insert basic drawings.