Face Recognition model using Local Binary
a) Serial Implementation b) Parallel Implementation using OpenMP( with HPC ).
Comparison of Both Models and proving which one is faster.
Training Step : For each image in the dataset, the following steps are applied:
▹ I used each 3 x 3 window in the image is processed to extract an LBP code.
▹ For each pixel in an image ,comparison with its 8 neighbours (on its left-top,
left-middle, left-bottom, right-top, etc.).
▹ Where the centre pixel’s value is less than the neighbour’s value, “1”.
▹ These values are converted into a decimal value and we will get a LBP code
▹ Histogram is computed, over the pixels, of the frequency of each “number”
occurring (i.e., each combination of which pixels are smaller and which are
greater than the centre).
Testing step : Histogram for the test image are generated as described previously.
• Distance values between test image’s histogram and training histograms
• Closest training histogram to determine the image is selected.
• Results are displayed.
This project effeciently depicts that the parallel implementation using OpenMp of the LBP Algorithm(or any) decreases the processing time significantly as compared to its serial implementation.