Shape Recognition Based on Projected Edges and Global Statistical Features

Pseudocode 1

function EPPSED(S)

N=64

% preprocessing

rotate(S, -orientation(S))

resize(S, [N, N], fit)

if horizontal_mass_center(S) > N/2 then

rotate(S, 180)

directions  ≔  []

% generate edge maps (EM) for every direction

for dir in directions

EM(dir)  ≔  convolution2d(S, EV(dir))

% for every location threshold the edge maps

for i in [1..N], j in [1..N]

for dir in directions

  ≔  

EMT(dir)  ≔  (EM(dir), )

% project thresholded edge maps and scale them

for dir in directions

PR(dir)  ≔  histogram(EMT(dir))

scale(PR(dir), N/4)

EPPSED  ≔  [PR(), PR(), PR(), PR()]

end