Opencv learning notes - linear filtering: box filtering, mean filtering, Gaussian filtering

Smoothing （ blurred ）

Generally used Reduce noise or distortion on the image . When reducing the image resolution , Use smoothing .

Eliminating noise components in an image is called Image smoothing or Filtering operation .

The purpose of image filtering is ：

        1、 Extract the feature of the object as the feature pattern of image recognition ;

        2、 In order to meet the requirements of image processing , Eliminate the noise in image digitization .

Smooth filtering It's low-frequency enhanced spatial filtering technology . It has two purposes :

1、 Fuzzy ;  2、 Eliminate noise .

Common image smoothing operation methods ：

linear ：

Box filtering ---------BoxBlur function

Mean filtering （ Neighborhood average filtering ）---------Blur function

Gauss filtering ---------GaussianBlur function

nonlinear ：

median filtering ---------medianBlur function

Bilateral filtering ---------bilateralFilter function

Linear filter

Linear filters are often used to eliminate unwanted frequencies in input signals or to select a desired frequency from many frequencies .

Common linear filters are as follows ：

         low pass filter ： Allow low frequencies to pass through ;

         High pass filter ： Allow high frequency passage ;

         Bandpass filter ： Allow a certain range of frequencies to pass ;

         Band stop filter ： Prevent a certain range of frequencies from passing and allow other frequencies to pass ;

         All pass filter ： Allow all frequencies to pass through , Just change the phase relationship ;

         Notch filter ： Prevent a narrow frequency range from passing through , It is a special band stop filter .

Box filtering

boxBlur function     sources\modules\imgproc\src Under the smooth.cpp

void boxFilter(InputArray src, OutputArray dst, int ddepth, 
               Size ksize, Point anchor = Point(-1,-1), 
               bool normalize = true, int borderType = BORDER_DEFAULT )

Parameters 1,src, The input image , Source image , fill Mat Class object . This function is independent of the channel , And can                 Process pictures with any number of channels . Be careful , The depth of the image to be processed should be CV_8U、CV_16U、                        CV_16S、CV_32F as well as CV_64F One of .

Parameters 2,dst, The target image , Need the same size and type as the source image .

Parameters 3,ddepth, The depth of the output image ,-1 Represents the depth of the original image , namely src.depth().

Parameters 4,ksize, The size of the kernel . It's usually used Size(w, h) To represent the size of the kernel , among w For pixel width ,h by                   Pixel height .Size(3, 3) It means 3 x 3 The nuclear size of ,Size(5, 5） It means 5 x 5 The nuclear size of .

Parameters 5,anchor, Represents an anchor （ The point that is smoothed ). The default value is Point(-1,-1). If the coordinate of this point is negative               If it's worth it , It means that the center of the core is the anchor point , So the default value is Point(-1,-1) Indicates that this anchor is in the core                   heart .

Parameters 6,normalize, The default value is true, Indicates whether the kernel is normalized by its region (normalized)

Parameters 7,borderType, Some kind of boundary pattern used to infer pixels outside the image . The default value is :                                                 BORDER_DEFAULT, Generally not used .

boxFilter The kernel representation of the function ：

among

According to the effect ：

 ksize The bigger it is , The more blurred the picture is .

Code ：

	Mat srcImg1;
	Mat srcImg2;

	srcImg1 = imread("E:\\img\\logo3.png");
	imshow(" Source graph srcImg1", srcImg1);

	boxFilter(srcImg1, srcImg2, -1, Size(14, 18));
	imshow(" Box filtering srcImg2", srcImg2);

Mean filtering

Each pixel of the output image of the mean filter is the average value of the corresponding pixel of the input image in the kernel window ( All pixels are weighted equally ), That is, normalized block filtering .

Mean filtering is A typical linear filtering algorithm . The main method is Neighborhood averaging , That is to use the average value of each pixel in an image area to replace the value of each pixel in the original image .

defects ：

        Cannot protect image details well , While denoising the image, it also destroys the details of the image , Make the image blurred , It can't remove the noise very well .

void blur(InputArray src, outputArraydst, Size ksize, 
          Point anchor = Point(-1, -1), int borderType = BORDER_DEFAULT)

Parameters 1, The input image , Source image ,Mat Class object . This function is independent of the channel , And can handle any               Picture of the number of channels . Be careful , The depth of the image to be processed should be CV_8U、CV_16U、CV_16S、                       CV_32F as well as CV_64F One of .

Parameters 2, Target image , Need the same size and type as the source image . It can be used Mat::Clone, Take the source image as the model                   plate , To initialize the target graph .

Parameters 3, The size of the kernel . It's usually used Size(w, h) To represent the size of the kernel , among w For pixel width ,h Is pixel high                     degree .Size(3,3) It means 3x3 The nuclear size of ,Size(5,5） It means 5x5 The nuclear size of .

Parameters 4, Represents an anchor （ The point that is smoothed ). The default value is Point(-1,-1). If the coordinates of this point are negative ,                It means that the center of the core is the anchor point , So the default value is Point(-1,-1) It means that the anchor is in the center of the core .

Parameters 5, Some kind of boundary pattern used to infer pixels outside the image . The default value is BORDER_DEFAULT, Generally do not make                     use .

According to the effect ：

Code ：

Mat srcImg1;
Mat srcImg2;

srcImg1 = imread("E:\\img\\logo3.png");
imshow(" Source graph srcImg1", srcImg1);

blur(srcImg1, srcImg2, Size(18, 18));
imshow(" Mean filtering srcImg2", srcImg2);

Gauss filtering

Gaussian filtering refers to the filtering operation using Gaussian function as the filtering function ;

Application scenarios ：      Noise reduction process of image processing .

The specific implementation of Gaussian filtering :

Use a template （ Convolution 、 Mask ） Scan every pixel in the image , Using the weighted average gray value of the pixels in the neighborhood determined by the template to replace the value of the central pixel of the template .

Gaussian blur of image is the convolution of image and normal distribution .

Gaussian blur is a Gaussian low-pass filter .

GaussianBlur function ： Blur an image with a Gaussian filter

Convolute the source image with the specified Gaussian kernel function , And support local filtering .

void GaussianBlur(InputArray src, outputArray dst, Size ksize, 
                  double sigmaX, double sigmaY = 0, 
				  int borderType = BORDER_DEFAULT )

Parameters 1, The input image , fill Mat Class object . It can be a single picture with any number of channels ,

            Be careful , The picture depth should be CV_8U、CV_16U、CV_16S、CV_32F as well as CV_64F One of .

Parameters 2, Target image , Need the same size and type as the source image . It can be used Mat::Clone, Take the source image as the model                   plate , To initialize the target graph .

Parameters 3, The size of the Gaussian kernel . among ksize.width and ksize.height Can be different , Must be positive and odd ,

              Or zero , All by sigma Come by calculation .

Parameters 4, Denotes that the Gaussian kernel function is in X The standard deviation of the direction .

Parameters 5, Denotes that the Gaussian kernel function is in Y The standard deviation of the direction . if sigmaY zero , Set it to sigmaX;

              If sigmaX and sigmaY All are 0, from ksize.width and ksize.height Work it out .

Parameters 6, Some kind of boundary pattern used to infer pixels outside the image . The default value is BORDER_DEFAULT, Generally do not make                     use .

According to the effect ：

 

Code ：

Mat srcImg1;
Mat srcImg2;

srcImg1 = imread("E:\\img\\logo3.png");
imshow(" Source graph srcImg1", srcImg1);

GaussianBlur(srcImg1, srcImg2, Size(5, 5), 0,  0);
imshow(" Gauss filtering srcImg2", srcImg2);