def testWlcgp(img2, numx, numy, base, E, P, nTrain, NumPerClassTrain):
img2 = ((img2 - np.mean(img2)) + 128) / np.std(img2) * 20
iiCell = commons.block(numx, numy, img2)
lbpII = []
lbpII = np.asarray(lbpII)
for k in range(numx):
for m in range(numy):
iiCellBlock = iiCell[k, m]
blockLBPII = wlcgpAlgorithm.wlcgp(iiCellBlock)
blockLBPII = np.transpose(blockLBPII)
if (m == 0) & (k == 0):
lbpII = blockLBPII
else:
lbpII = np.concatenate((lbpII, blockLBPII))
# end
# end
lbpII = lbpII.reshape(1, lbpII.shape[0])
tcoor = lbpII.dot(base)
tcoor = np.transpose(E).dot(np.transpose(tcoor))
mdist = [None] * nTrain
for k in range(nTrain):
mdist[k] = np.linalg.norm(tcoor - P[:, k])
# end
# 3 NN algorithm
index2 = np.argsort(mdist)
dist = np.sort(mdist)
class1 = int(np.math.floor(index2[1] / NumPerClassTrain - 0.1) + 2)
class2 = int(np.math.floor(index2[2] / NumPerClassTrain - 0.1) + 2)
class3 = int(np.math.floor(index2[3] / NumPerClassTrain - 0.1) + 2)
result = -1
if (class1 != class2) & (class2 != class3):
result = class1
else:
if class1 == class2:
result = class1
if class2 == class3:
result = class2
# end
return result
NumPerClassTest = NumPerClass - NumPerClassTrain #
allsamples = [] #
numx = 4 # image segmentation settings
numy = 2
filePath = filepath + "\s" + str(1) + "\\" + str(1) + ".pgm" # filepath to read
img = cv2.imread(filePath, 0)
xsize, ysize = img.shape # get file size xsize - pionowo; ysize - poziomo
img = np.double(img)
img = ((img - np.mean(img))+128)/np.std(img) * 20 # ok
iCell = commons.block(numx, numy, img)
lbpI = []
k = 0
# while k <= numx:
for k in range(1): #numx
#k += 1
#m = 0
#while m <= numy:
for m in range(1): #numy
#m += 1
iCellBlock = iCell[k,m] #k,m
# iCellBlock = #cell2mat --nie wiem czy potrzebujemy bo ta funkcja jest specyficzna dla matlaba i rozpisuje "cell" na
blockLBPI = wlcgpFile.wlcgp(iCellBlock) #!!!!