def addOneRow(npimg,npgray,npnew, pos, gdratio):
npimgt = transposeImg(npimg)
npnewt = transposeImg(npnew)
npgrayt = transposeGray(npgray)
post = transposeGray(pos)
ret, retnew, retpos = addOneColumn(npimgt, npgrayt, npnewt, post, gdratio)
return transposeImg(ret), transposeImg(retnew), transposeGray(retpos)
def deepAddOneRow(npimg,npgray,npnew, pos, gdratio,heat):
npimgt = transposeImg(npimg)
npnewt = transposeImg(npnew)
npgrayt = transposeGray(npgray)
heatt = transposeGray(heat)
post = transposeGray(pos)
ret, retnew, retpos = deepAddOneColumn(npimgt, npgrayt, npnewt, post, gdratio, heatt)
return transposeImg(ret), transposeImg(retnew), transposeGray(retpos)
def deepDeleteOneRow(npimg, npgray, gdratio, heat):
npimgt = transposeImg(npimg)
npgrayt = transposeGray(npgray)
heat = transposeGray(heat)
return transposeImg(deepDeleteOneColumn(npimgt, npgrayt, gdratio, heat))
def deleteOneRowForward(npimg, npgray, gdratio):
npimgt = transposeImg(npimg)
npgrayt = transposeGray(npgray)
return transposeImg(deleteOneColumnForward(npimgt, npgrayt, gdratio))
def retargetOptimal(npimg, gdratio, r, c):
origr, origc, _ = npimg.shape
rdelete = origr - r
cdelete = origc - c
T = np.zeros((rdelete + 1, cdelete + 1))
# 1 means delete col
I = np.zeros((rdelete + 1, cdelete + 1), dtype=np.int8)
savedImgs = []
savedGrays = []
savedImgs.append(npimg)
for i in range(rdelete + 1):
print(i)
for j in range(cdelete + 1):
if i == 0:
if j == 0:
savedGrays.append(npimg2npgray(npimg))
continue
else:
npimg = savedImgs[-1]
npgray = npimg2npgray(savedImgs[-1])
energy, lastDir = generateColumn(computeEnergy(npimg, npgray, gdratio))
minVal = np.min(energy[-1])
T[i,j] = T[i,j-1] + minVal
I[i,j] = 1
newimg = deleteOneColumnWithEnergyProvided(npimg,energy,lastDir)
savedImgs.append(newimg)
savedGrays.append(npimg2npgray(newimg))
else:
if j == 0:
npimg = savedImgs[0]
npgray = savedGrays[0]
npimgt = transposeImg(npimg)
npgrayt = transposeGray(npgray)
energy, lastDir = generateColumn(computeEnergy(npimgt, npgrayt, gdratio))
minVal = np.min(energy[-1])
T[i,j] = T[i - 1,j] + minVal
newimgt = deleteOneColumnWithEnergyProvided(npimgt, energy, lastDir)
newimg = transposeImg(newimgt)
savedImgs[0] = newimg
savedGrays[0] = npimg2npgray(newimg)
else:
# col
npimgCol = savedImgs[j - 1]
npgrayCol = savedGrays[j - 1]
energyCol, lastDirCol = generateColumn(computeEnergy(npimgCol, npgrayCol, gdratio))
minValCol = np.min(energyCol[-1])
# row
npimgRow = savedImgs[j]
npgrayRow = savedGrays[j]
npimgRowt = transposeImg(npimgRow)
npgrayRowt = transposeGray(npgrayRow)
energyRow, lastDirRow = generateColumn(computeEnergy(npimgRowt, npgrayRowt, gdratio))
minValRow = np.min(energyRow[-1])
if T[i - 1,j] + minValRow < T[i, j - 1] + minValCol:
newimgt = deleteOneColumnWithEnergyProvided(npimgRowt, energyRow, lastDirRow)
newimg = transposeImg(newimgt)
else:
newimg = deleteOneColumnWithEnergyProvided(npimgCol, energyCol, lastDirCol)
I[i, j] = 1
savedImgs[j] = newimg
savedGrays[j] = npimg2npgray(newimg)
for i in range(rdelete + 1):
for j in range(cdelete + 1):
print(I[i,j],end=' ')
print()
return savedImgs[-1]
