我使用以下步骤在cv2.dft()
中使用python 3.5
。
N=32
Icos = np.zeros((32,32))
for i in range(0,N):
for j in range(0,N):
myPi = 2*math.pi/N
th1 = 8*i + 6*j
th2 = 4*i + 2*j
th3 = 2*j
Icos[i,j] = (0.25 * math.cos(myPi*th1) + 0.75*math.cos(myPi*th2) + math.cos(myPi*th3))
image = Icos
dft = cv2.dft(np.float32(image),flags =cv2.DFT_COMPLEX_OUTPUT)
#Shift to the center
dft_shift = np.fft.fftshift(dft)
#Magnitude Calculation
ms = np.log(1+cv2.magnitude(dft_shift[:,:,0],dft_shift[:,:,1]))
mask_out = 'Mask LowPass'
mask = np.zeros((rows,cols,2),np.uint8)
for i in range(0,rows):
for j in range(0,cols):
point = math.sqrt(math.pow((i-crow),2)+math.pow((j-ccol),2))
if point <= radius1:
mask[i,j]=1
#apply mask
fshift = dft_shift*mask
#Inverse
f_ishift = np.fft.ifftshift(fshift)
img_back = cv2.idft(f_ishift)
img_back = cv2.magnitude(img_back[:,:,0],img_back[:,:,1])
#Fourier Transform of the resulted image
dft_res = cv2.dft(np.float32(img_back),flags =cv2.DFT_COMPLEX_OUTPUT)
#Shift to the center
dft_shift_res = np.fft.fftshift(dft_res)
#Magnitude Calculation
ms_res = np.log(1+cv2.magnitude(dft_shift_res[:,:,0],dft_shift_res[:,:,1]))
如下所示,当我使用radius1
= 7时,所得到的反转图像的幅度谱似乎不正确。
反转图像的幅度谱(如果我再次执行dft / fftshift)不会与此类似吗?
反转图像看起来应该是这样吗?