how to convert video in to image frames using matlab?

129 次查看（过去 30 天）

显示更早的评论

Akhil krishnan 2013-1-7

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/58111-how-to-convert-video-in-to-image-frames-using-matlab

评论： Image Analyst 2021-5-10

how to convert video in to image frames using matlab?

5 个评论
显示 3更早的评论隐藏 3更早的评论

Walter Roberson 2019-4-6

Rose Mohammed:

"i", "p", "b" frames are only used by some video codecs.

The code doing the encoding of frames compares the current frame to a previous frame (and possibly a following frame) and decides how much change there has been. If it decides that there has been so much change that describing the changes would take more space than just sending the complete frame, then it sends the complete frame, marking it in the file as an "i" frame. There is not necessarily anything special in terms of content for the frame: it would simply have changed enough relative to the previous one to be hard to describe the changes in a compact form.

If a frame has mostly the same content as a previous frame, then sometimes it can be more efficient to describe the changes. For example if someone blinks, then you might only have to send enough information to describe the pixels where the eyes are, with everything else the same. Frames that can be described efficiently in terms of the content of the frame before are encoded into the file as "p" frames that just describe the changes to the frame "previous".

Some video standards require sending "i" frames at fixed intervals even if not much has changed. That makes it practical to seek into the video stream: instead of having to find an "i" frame and decode a possibly long list of changes, you can just advance every so-many frames (usually 6) to find a frame close to where you want to start, and decode the little bit from there.

But if you place "i" frames at fixed intervals then because a big change is not always right at that interval, you might want to refer to the next "i" frame and describe changes relative to it instead of to the previous "i" frame. Or you might want to use information from both, using a "b" (both) frame that describes changes from the previous and next frame. Again, there is not necessarily anything special about the frames that lead to this, and can simply be due to accumulated comparatively small changes.

If you think about this for a moment, you will realize that the kind of encoding used for any given frame depends upon what has happened in previous frames more than what is in the frame itself. If you only have one frame in front of you, you would have to encode it as an "i" frame. If you have several frames, then how you encode this one depends on how much change there is relative to previous frames.

The next thing you need to know is that all of the video readers provided by Mathworks will only return to you fully decoded frames. None of them will tell you how any given frame was encoded in the video file or video stream. And, as I indicated above, you cannot tell just by looking at the frame itself.

In order to find out how any given frame was encoded in the file or stream, you would need to call upon third-party decoding libraries and ask them to give you information about the file or stream. Some of the libraries you could investigate are "libmpeg" and "libmp4".

Walter Roberson 2019-4-11

H.264 is generally adaptive to determine GOP structure.

That said, there is also choice in the matter. For some purposes, the exact structure can be important and GOP that follow a fixed pattern can be generated. There are tradeoffs to be made for bitrate considerations (how fast does your media read information?); or for CPU speed considerations (slower CPUs are less expensive and might produce less heat, but require less complicated video streams); or for how quickly you can seek to arbitrary points (e.g. chapter search or user fast forward or (more difficult) reverse)

https://www.researchgate.net/figure/Example-of-fixed-and-variable-GOP-structures-Efficient-video-coding-techniques-based-on_fig1_221126584

Rose Mohammed 2019-4-12

Thank you very much prof

请先登录，再进行评论。

请先登录，再回答此问题。

回答（3 个）

Lalit Patil 2013-1-7

4
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/58111-how-to-convert-video-in-to-image-frames-using-matlab#answer_70344

在 MATLAB Online 中打开

 vid=mmreader('video.avi');
 numFrames = vid.NumberOfFrames;
 n=numFrames;
 for i = 1:2:n
 frames = read(vid,i);
 imwrite(frames,['Image' int2str(i), '.jpg']);
 im(i)=image(frames);
 end

14 个评论
显示 12更早的评论隐藏 12更早的评论

BHANU SRINIVASA 2020-1-27

can i get this mmreader file

im not able to find on the net

Walter Roberson 2020-1-27

mmreader was removed sometime around 2011. You can use VideoReader to replace it in most cases.

请先登录，再进行评论。

Image Analyst 2016-3-17

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/58111-how-to-convert-video-in-to-image-frames-using-matlab#answer_213992

ExtractMovieFrames.m

Namith:

You can use the VideoReader class to set the starting time of the video, then use read() to extract a frame from the video at exactly that time. Then use imwrite() to save it to disk. Attached is a demo that does something similar (but not exactly what you want). The demo writes out every single frame to files. You just need to set the time with the method for doing that before you extract the frame.

7 个评论
显示 5更早的评论隐藏 5更早的评论

Randall Ang 2021-5-10

Hi, I tried using the demo that you provided, and it worked well. However, is there a way to ensure that the aspect ratio of the images remain the same as the video?