[image fusion] multi focus image fusion in DCT domain based on MATLAB [including Matlab source code, 1973]

One 、 be based on DCT Introduction of image fusion algorithm based on transform

In the process of image fusion , The most important thing is how to extract low-frequency coefficients and fusion criteria of low-frequency coefficients . be based on DCT The principle of the transformed image fusion algorithm is shown in the figure 2 Shown .

chart 2 DCT Principle of fusion algorithm

The algorithm steps are as follows .
step 1 Accurately register the source image to be fused .
step 2 The size of each piece participating in the fusion is M×N The source images of are divided into m×m Small pieces .
step 3 For steps 2 All the small pieces obtained in DCT Transformation .
step 4 Yes DCT The transformed blocks extract low-frequency coefficients and high-frequency coefficients respectively , The low and high frequency coefficients at the corresponding positions are fused according to the low and high frequency fusion criteria .
step 5 For fused DCT The coefficient is calculated IDCT, Finally, we get the fused image .

2.1 Low frequency coefficient fusion
The most important part of information for vision , All focus on the low frequency of the image . Low frequencies represent slow changes between image pixels , That is, the image frame part [11] . In order to maintain the visibility of the image , Keep the low-frequency part of the image , The change of low-frequency part may cause great changes in the image . be based on DCT The low-frequency coefficients of the transformed fused image are averaged , Suppose there is p Multiple exposure images , Can be defined as

In style Gk(i,j) Is the source image via DCT Low frequency coefficient extracted after transformation ;G(i,j) Is the low-frequency coefficient after fusion ;wk It's the weighting factor .

2.2 High frequency coefficient fusion
The high frequency coefficient corresponds to the detail information of the image , Features such as edges . The fusion rules of high-frequency coefficients are as follows .
(1) The segmented image is DCT After transformation , Extract its high-frequency coefficient .
Calculate the high-frequency coefficients respectively D(i,j) In pixels (i,j) Centred (2k+1)×(2k+1) The expression of image standard deviation in the neighborhood is


(2) remember p The regional standard deviations of a high-frequency coefficient of multiple exposure images are [C1(i,j),C2(i,j),…,Cp(i,j)], Then the weight coefficient corresponding to the extracted high-frequency coefficient is

(3) By the type (7) We can draw p Weight of multiple exposure images , Compare them , High frequency coefficient after fusion D(i,j) Is the high frequency coefficient corresponding to the maximum weight coefficient . If

that

Two 、 Partial source code

clc
clear
close all

%Select First Image
disp('Please Select First Image:')
[filename, pathname]= uigetfile({
    '*.jpg;*.png;*.tif'},'Select First Image');
path=fullfile(pathname, filename);
im1=imread(path);
disp('Great! First Image is selected')

%Select Second Image
disp('Please Select Second Image:')
[filename, pathname]= uigetfile({
    '*.jpg;*.png;*.tif'},'Select Second Image');
path=fullfile(pathname, filename);
im2=imread(path);
disp('Great! Second Image is selected')


if size(im1,3) == 3     % Check if the images are grayscale
    im1 = rgb2gray(im1);
end
if size(im2,3) == 3
    im2 = rgb2gray(im2);
end

if size(im1) ~= size(im2)	% Check if the input images are of the same size
    error('Size of the source images must be the same!')
end

3、 ... and 、 Running results

Four 、matlab Edition and references

1 matlab edition
2014a

2 reference
[1] Liu Weihua , Mayanghua , Liu Ying . be based on DCT Transform multi exposure image fusion method [J]. Journal of Xi'an University of Posts and telecommunications . 2016,21(06)

3 remarks
This part of the introduction is taken from the Internet , For reference only , If infringement , Contact deletion