mpeg file

Hi

I have two task that are similar that I dont know how to do:

1 - I want to be able to import an mpeg video and save each frame of this movie as individual images into a folder.

2 - Also, I want to be able to create an mpeg video given a number of individual images saved in a folder.

Answer 1

[Walter Roberson]

If you have a new enough MATLAB version, you could perhaps us the VideoReader class; older versions called it the mmreader. Extract frame by frame and use imwrite(). See http://matlab.wikia.com/wiki/FAQ#How_can_I_process_a_sequence_of_files.3F for information on automatic filename generation for such purposes.

Given a series of frames read in by imread() (in conjunction with dir() to find the file names), you can use the VideoWriter() class to produce a .avi .

Answer 2

[James]

Sorry I need further help...

For task 1: - I have imported the mpeg video using:

object = mmreader('train.mpg')

and I know how to scroll through each frame using:

for i=1:NumberOfFrames
read(object,i)
end

But how do I save this frame as a jpg?

[Walter Roberson]

for i=1:NumberOfFrames

thisframe = read(object,i)

imwrite(thisframe, AppropriateFilename, 'jpg')

end

For more information on how to construct the AppropriateFilename, see the FAQ I referenced earlier.

[James]

for i=1:10

thisframe=read(object,i)

thisframe=rgb2gray(thisframe)

filename-strcat('frame',num2str(i),'.jpg')

imwrite(thisframe,filename,'jpg')

end

It creates the first image (frame1), however I then get the following error:

??? Error using ==> minus

Matrix dimensions must agree.

[Walter Roberson]

filename-strcat('frame',num2str(i),'.jpg')

is a request to *subtract* something from filename. You want assignment.

[James]

oops absolute school boy error! Cheers.

[James]

I was just wondering....

I need to really put the frame into double format so I can do some calculations.

When I put this into my code as shown below, then the image appears differently. Why is this? And how do I change it back?

for i=1:10

thisframe=read(object,i)

thisframe=rgb2gray(thisframe)

thisframe=double(thisframe)

filename=strcat('frame',num2str(i),'.jpg')

imwrite(thisframe,filename,'jpg')

end

[Walter Roberson]

See im2double() http://www.mathworks.com/help/toolbox/images/ref/im2double.html

[James]

Sorry i dont really understand this and I dont really know why I have to convert it to double to be able to perform calculations on the frame?!

[James]

It seems that im2double() worked though instead of double().

But in my previous work I used double() as shown below. Does this mean what I have done below is wrong?

for k = 1 : nFrames

mov(k).cdata = read(videoobject, k);

end

Img = double(Video(1,n).cdata);

[Walter Roberson]

Using double() is not necessarily wrong, but you have to keep track of the fact that you used it and you either have to convert back to the original class later before displaying (or writing) the image, or else you have to do the appropriate rescaling later.

Data arrays of class double are presumed to represent color values as a numeric value in the range 0 to 1 (inclusive). Data read via imread() or mmreader is almost always an unsigned integer data type such as uint8 or uint16, which represent colors as integers in the range 0 to 255, or 0 to 65535 (respectively). If you double() something in the range 0 to 255 or 0 to 65535, the double precision value will mostly be outside of the expected value for double precision color representation, 0 to 1, and the graphics subsystem will convert the 0.0 to 255.0 by clamping everything about 1.0 to be 1.0, if you ever try to display it or write it.

im2double() takes care of looking at the source data type and if necessary scales the data, such as by doing double(I)./255 so that the data comes out in the range 0 to 1, suitable for display and writing purposes.

[James]

Hi

It seems when I double()a frame I still have the same numbers, which is not what you are saying?! However, when I use im2double() then I do get numbers between 0 and 1.

[Walter Roberson]

Yes, double() keeps the same numeric values, such as uint8(42) becoming the double precision number 42.0. That is fine for some kinds of computations, but if you try to display or imwrite() a double precision array with such values you will have problems, as the expected range for representing graphics in double precision variables is 0 to 1. im2double() converts the data type _and_ rescales the values to be in the range 0 to 1.