omaraskr1/parralel_background_subtraction

parralel_background_subtraction

: Parallel Background Subtraction

Description:

It’s a technique for removing static background by subtracting a set of images to obtain a final image with the objects only without a background. This is a basic technique which will work only on tiny/small motion changes in the images that are given.

Input and Output:

Steps: • Transform Images (NXN) to Vectors [(N2) X 1] and put them in large array [ N2 X M ] such that M = Number of Images

1. Estimated Background image B is the mean of the collection input streams of size M For example : B0 = ( ( pixel 0 in image 1 + pixel 0 in image 2 +........ ) / M) B1 = ( ( pixel 1 in image 1 + pixel 1 in image 2 +........ ) / M)

2. Foreground Mask image X is check of subtract mean from image based on a threshold
   For example : X0_new = | B0 - X0_new_input_frame | > TH X1_new = | B1 - X1_new_input_frame | > TH *TH is a threshold which you can tune (dependent on your visual results) • Render your results • Test your code with these different conditions: 1- different image sizes (e.g if the image size is N * N, so test your code if the image size is 5N * 5N and 10N * 10N) • Record your Results

   N* N 5N* 5N 10N* 10N

sequential code MPI Sol1 MPI Sol (bonus) Deliverables • The source code • A report on the findings and enhancements made to performance. The project discussion will be individual Check these videos: ● Basic Background Subtraction - Motivation ● Background Mean Averaging