今天写的,主要有:
(1)正常的图片,故障的图片,先求LBP特征。因为LBP有光照不变性和旋转不变性
(2)把图片分成上下两部分,这样可以使故障的部分显得更显著
(3)融合PSRN和SSIM两种方法,这样会有更好的鲁邦性
"""
"""
import os
import cv2
import time
import numpy as np
from skimage.transform import rotate
from skimage.feature import local_binary_pattern
from skimage import data, io
from skimage.color import label2rgb
import skimage
import math
####
def compute_psnr(img1, img2):
if isinstance(img1,str):
img1=io.imread(img1)
if isinstance(img2,str):
img2=io.imread(img2)
mse = np.mean( (img1/255. - img2/255.) ** 2 )
if mse < 1.0e-10:
return 1000000000000
PIXEL_MAX = 1
psnr = 20 * math.log10(PIXEL_MAX / math.sqrt(mse))
return mse, psnr
def ssim(img1, img2):
C1 = (0.01 * 255) ** 2
C2 = (0.03 * 255) ** 2
img1 = img1.astype(np.float64)
img2 = img2.astype(np.float64)
# kernel = cv2.getGaussianKernel(11, 1.5)
kernel = cv2.getGaussianKernel(11, 1.5)
window = np.outer(kernel, kernel.transpose())
mu1 = cv2.filter2D(img1, -1, window)[5:-5, 5:-5] # valid
mu2 = cv2.filter2D(img2, -1, window)[5:-5, 5:-5]
mu1_sq = mu1 ** 2
mu2_sq = mu2 ** 2
mu1_mu2 = mu1 * mu2
sigma1_sq = cv2.filter2D(img1 ** 2, -1, window)[5:-5, 5:-5] - mu1_sq
sigma2_sq = cv2.filter2D(img2 ** 2, -1, window)[5:-5, 5:-5] - mu2_sq
sigma12 =