What do you mean by Key frames in video sequence?

If you are looking for a distinctive frame which has special information and parameters, you should describe it first. what do you want?

Key-frames are defined as the representative frames of a video stream, the frames that provide the most accurate and compact summary of the video content. You understand that by default the identification of key-frames is an ill-defined problem. The "correct" key-frame detection process depends not only on the application but also on the personal "definition" of the user/developer of what a video summary should include.

As a result, there are not some commonly accepted key-frame estimation techniques.

Some hints-ideas:

1. Most key-frame extraction techniques start with the decomposition of the video into temporal segments (shots or scenes) and then extract a fixed or not number of key-frames per temporal segment. In some applications even the selection of the middle frame of each shot may be a decent approach for key-frame estimation (the middle, not the first or the last because at the beginning and at the end of shots or scenes may be frames belonging to artistic passages from one shot to another, i.e. visual effects, such as fade in/fade out).

2. A widely used approach is to use the low-level visual information of all video frames (or all shot frames or all scene frames) in order to group them using e.g. k-means and then select as key-frames the ones that are more similar to the groups centers of the groups.

3. Another class of approaches employ a sequential search to video stream. Such techniques start by a "root" key-frame (usually randomly selected as one of the first frames of the video) and then compare one by one the next frames, until a frame with significantly different low-level visual content is found. Then, this becomes the "root" key-frame and the process continues from the next video frame.

The above are just a small sub-set of the techniques used for key-frame extraction. There is an extensive literature on the subject and several review papers. What I would suggest you is to take a look, having always in mind that there is no "absolutely" correct answer on the key-frame estimation problem, but all depend on the application needs/constraints/objectives.

The key frames define the starting and ending points of any transition. A sequence of keyframes defines which movement the viewer will see, whereas the position of the keyframes on the video or animation defines the timing of the movement. Because only two or three keyframes over the span of a second do not create the illusion of movement, the remaining frames are filled with inbetweens.

When keyfarems are identiied then only thse can be used for procesing the video, so that time is saved. With hother words key frames in the video carries the information required.

Some good methods for key frame detection in video sequence:

"Video key frame extraction through dynamic Delaunay clustering with a structural constraint"

This one is an old method, but it is one of the most cited in scientific literature:

http://ieeexplore.ieee.org/xpls/abs_all.jsp?arnumber=543588&tag=1

This one is based on a generative model.

http://pdf.aminer.org/000/293/568/an_em_algorithm_for_video_summarizationgenerative_model_approach.pdf

There are some methods based on visual saliency information:

http://dl.acm.org/citation.cfm?id=641116

As I said, it depends on the application. In your case (is it a medical type of video?), I would suggest you to use mpeg compression and keep only the I-frames. I remind you that mpeg compression keeps only the frames that can't be reconstructed by neighboring frames (as I-frames) and predicts every other frame using these. Have a look here:

http://vsr.informatik.tu-chemnitz.de/~jan/MPEG/HTML/mpeg_tech.html

http://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol2/ck4/article2.html

I think that there are available tools on the web that retrieve only I-frames from mpeg compressed videos.

Thak you

Hi Christos,

I would recommend GIST (http://people.csail.mit.edu/torralba/code/spatialenvelope/) for global level search of key frames in a video.

Tauseef

An easy way is to calculate the summation of the differences between pixels in each frame, if it is above specific threshold identify as a key frame.

hi

go through the following papers , you may got some idea on this questions

http://research.microsoft.com/en-us/um/people/yongrui/ps/icip98.pdf

http://grail.cs.washington.edu/projects/control/smokeControl.pdf

https://www.google.co.in/url?sa=t&rct=j&q=&esrc=s&source=web&cd=6&cad=rja&uact=8&ved=0ahUKEwi02v65ru_LAhWCBo4KHd8CAdAQFghIMAU&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D6112830&usg=AFQjCNG-9XDDo_VLOcpX_bdXs0L6CyZ2Dw&bvm=bv.118443451,d.c2E

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5497695&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5497695

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5957780&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D5957780

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1227695&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D1227695

http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=1000355&url=http%3A%2F%2Fieeexplore.ieee.org%2Fxpls%2Fabs_all.jsp%3Farnumber%3D1000355