[Learning OpenCV 4] Converting and Normalizing Image Pixel Data Types

I. What Is Normalization

  Normalization is about constraining the data to be processed, after processing (via some algorithm), within a certain range. This makes later data processing easier and also helps the program converge faster during runtime.   The purpose of normalization is to make incomparable data comparable while preserving the relative relationship between the two data items being compared, such as their ordering; or, for plotting, data that was originally hard to plot on one chart can be normalized so that relative positions on the chart are easy to show.

II. Normalization Methods

2.1 Basic API

void normalize(
	InputArray 			src,
	InputOutputArray 	dst, 
	double 		alpha = 1, 
	double 		beta = 0, 
	int 		norm_type = NORM_L2, 
	int 		dtype = -1, 
	InputArray 			mask = noArray()
);

Parameter	Purpose
src	Input array
dst	Output array
alpha	Normalized minimum value
beta	Normalized maximum value
norm_type	Normalization type
dtype	When negative, the output array type matches the input array type
mask	Indicator function for whether the operation applies only to specified elements

norm_type has the following types:

• NORM_MINMAX: Array values are translated or scaled to a specified range; linear normalization, commonly used.
• NORM_INF: No definition found for this type; based on the corresponding entry in OpenCV 1, it may normalize the C-norm of the array (the maximum absolute value)
• NORM_L1 : Normalizes the L1 norm of the array (sum of absolute values)
• NORM_L2: Normalizes the (Euclidean) L2 norm of the array

2.2 Example Program

void MyDemo::normalize_Demo(Mat& image) {
	Mat dst;
	std::cout << image.type() << std::endl;	//CV_8UC3
	image.convertTo(image, CV_32F);			//像素数据转换为浮点数数据
	std::cout << image.type() << std::endl;	//CV_32FC3
	normalize(image, dst, 0, 1.0, NORM_MINMAX);	//归一化
	std::cout << dst.type() << std::endl;
	imshow("Normalize", dst);
}