how to detect shot boundary in video

SBD using adaptive threshold

Answer 1

[prabhat kumar sharma]

Hello Malathi,

Detecting shot boundaries in a video involves identifying transitions between different scenes or shots. One common approach is to use an adaptive threshold method. Here's a concise guide on how to achieve this:

1. Extract Frames:

• Convert the video into a series of frames.

2. Compute Frame Differences:

• Calculate the difference between consecutive frames. This can be done using pixel-wise subtraction or more sophisticated methods like histogram differences.

3. Adaptive Thresholding:

• Compute a threshold that adapts based on the frame differences. This can be done using statistical measures like mean and standard deviation of differences over a sliding window.

4. Detect Boundaries:

• Identify frames where the difference exceeds the adaptive threshold. These frames are likely to be shot boundaries.

Here is a basic example :

% Read video
video = VideoReader('video.mp4');
numFrames = video.NumFrames;

% Initialize variables
frameDiffs = zeros(numFrames-1, 1);

% Compute frame differences
for i = 1:numFrames-1
    frame1 = read(video, i);
    frame2 = read(video, i+1);
    frameDiffs(i) = sum(abs(frame1(:) - frame2(:))); % Sum of absolute differences
end

% Adaptive threshold
windowSize = 30; % Example window size
thresholds = movmean(frameDiffs, windowSize) + movstd(frameDiffs, windowSize);

% Detect shot boundaries
shotBoundaries = find(frameDiffs > thresholds);

% Display results
disp('Detected shot boundaries at frames:');
disp(shotBoundaries);

I hope it helps!