How to split a video into frames and analyze change in pixel intensities over time.

Hi all! I am currently working on a project using the raspberry pi camera board with MatLab. I currently have streamed h264 (1920 x 1080p) video from the pi and saved it, however, I need to split the video into the individual frames for analysis.

After splitting the video into frames, I would like to specify a 500x500 pixel section in EACH image and calculate the average pixelation intensity detected for each matrix of pixels. I would then need to be constantly storing this average pixel value as the video streams.

Have any of you worked with this before? Any ideas?

Thank you.

Answer 1

[Image Analyst]

That should be a very straightforward adaptation of my video processing tutorial, attached.

[Geoff Green]

Hey, Thanks for the tutorial. This is very helpful for me as well. I just wanted to know what the different plots (red, green, and blue) signified. Also, is the black line the main plot that we should be looking at as far as the true amount of change frame-by-frame? It seems to stick out and according to the code, I believe it was created separate of the other 3 if i'm not mistaken.

Additionally, what are the units on the Y-axis? Thanks in advance!

[Image Analyst]

As you can see in this line, which gets plotted as 'k-' (black):

meanGrayLevels(frame) = mean(grayImage(:));

the black line is simply the mean of all pixels. So it's the average of all red, all green, and all blue together. They're equally weighted so each has 1/3 weight. Of course each color is also plotted by itself so you can see the individual mean colors too.

[Geoff Green]

Oh, ok. That makes perfect sense. And what about the y-axis? What units should I associate that with?

[Image Analyst]

Gray levels.

[Geoff Green]

Thanks

[Klarizza Quilandrino]

How would you change the plot so that it’s pixel value vs. time (not frame)?

[Image Analyst]

You'd have to convert the frame index to a time. You need to find out the frame rate (maybe it's a property of the VideoReader object), and then you can figure out the time between frames and then multiply that by the index.

[Walter Roberson]

Ask the VideoReader object for its CurrentTime property. This also takes into account variable framerates.

Answer 2

[Abdullah Said]

i currently doing my final year project in detect suspicious activity in examination hall.Where i already detect skin color in video frames.My question how to detect movement of skin color in certain area that we already decided in video frames.thank you

[Image Analyst]

Check out papers here: http://www.visionbib.com/bibliography/contentsmotion-f.html#Motion%20--%20Feature-Based,%20Long%20Range,%20Motion%20and%20Structure%20Estimates,%20Tracking,%20Surveillance,%20Activities

You might use optical flow or other frame differencing methods to detect motion. Then, on the detected blobs, use color classification to determine if it's skin or note. See my File Exchange or see the attached demo.