1.标准化处理与归一化
1)白化处理(图像标准化处理);
图像标准化是将数据通过去均值实现中心化的处理,根据凸优化理论与数据概率分布相关知识,数据中心化符合数据分布规
律,更容易取得训练之后的泛化效果,数据标准化是数据预处理的常见方法之一。
2)归一化处理。
归一化不会改变图像本身的信息存储,取值范围从0~255已经转化为0~1之间。
import cv2
pic_path = 'D:/image/train/image/1.png'
def normalization(input):
pic = cv2.imread(input)
pic = pic.astype(np.float32)
# zero-center
pic -= [np.mean(pic[..., 0]), np.mean(pic[..., 1]), np.mean(pic[..., 2])]
# normalize
pic /= [np.std(pic[..., 0]), np.std(pic[..., 1]), np.std(pic[..., 2])]
print(pic[..., 0])
pic *= (pic > 0)
pic *= 100
pic = pic * (pic <= 255) + 255 * (pic > 255)
pic = pic.astype(np.uint8)
print(pic[..., 0])
cv2.imshow('', pic)
cv2.waitKey(0)
if __name__ == '__main__':
# 均值0中心化(zero-center), 规范化(normalize)
normalization(pic_path)
2.直方图与均衡化
直方图在图像处理中非常重要,利用图像直方图对对比度进行调整的方法。
基本思想:把原始图的直方图变换为均匀分布的形式,这样就增加了像素灰度值的动态范围,从而达到增强图像整体对比度的效果。
import numpy as np
from skimage import exposure, data
import cv2
import matplotlib.pyplot as plt
image = cv2.imread('D:/image/train/image/1.png')
#灰度直方图
#calcHist(images, channels, mask, histSize, ranges, hist=None, accumulate=None)
hist = cv2.calcHist([image],[0],None,[256],[0,256])
#灰度图像-灰度图片-直方图均衡化
hist2 = cv2.equalizeHist(image)
###彩色直方图
chans = cv2.split(image)
colors = ("b","g","r")
plt.figure()
plt.title("Flattened Color Histogram")
plt.xlabel("Bins")
plt.ylabel("# of Pixels")
for (chan,color) in zip(chans,colors):
hist = cv2.calcHist([chan],[0],None,[256],[0,256])
plt.plot(hist,color = color)
plt.xlim([0,256])
#彩色均衡化
def hisEqulColor(img):
ycrcb = cv2.cvtColor(img, cv2.COLOR_BGR2YCR_CB)
channels = cv2.split(ycrcb)
print (len(channels))
cv2.equalizeHist(channels[0], channels[0])
cv2.merge(channels, ycrcb)
cv2.cvtColor(ycrcb, cv2.COLOR_YCR_CB2BGR, img)
return img
image = cv2.imread('D:/image/train/image/1.png')
cv2.imshow('im1', image)
cv2.waitKey(0)
eq = hisEqulColor(image)
cv2.imshow('image2',eq )
cv2.waitKey(0)