Data Standardization——图像z-score normalization及python实现

z-score normalization就是标准化，将影像的R、G、B分别减去其均值然后再除以方差
在深度学习中，这样做可以加快模型的收敛速度。

imgdir = os.listdir(filepath)
mean = np.zeros(3)
std = np.zeros(3)
count = 0
for z in imgdir:
	count += 1
	img = cv2.imread(filepath + z)	
	for d in range(3):
		mean[d] += img[:, :, d].mean()
		std[d] += img[:, :, d].std()
mean = mean/count
std = mean/count
print (list(mean), list(std))