离散型哈尔小波变换简单实现

二阶哈尔变换矩阵如下
$$\left[\begin{array}{cc}\frac{1}{\sqrt{2}}& \frac{1}{\sqrt{2}}\\ \frac{1}{\sqrt{2}}& -\frac{1}{\sqrt{2}}\end{array}\right]$$
经过小波变换后，可以把数据划分为不同的成分：高频成分，低频成分。
多阶小波变换，对于低频成分可以进行划分。

一维离散型哈尔小波变换

实现代码

function [ A,D ] = haarLinearTranslation(data)
	ha=[0.7071,0.7071; 0.7071,-0.7071];
	[~,n]= size(data);
	data = reshape(data,2,n/2);
	data = ha*data;
	A = data(1,:);
	D = data(2,:);
end

调用

load noisdopp
[A,D]=haarLinearTranslation(noisdopp);
subplot(3,1,1)
plot(noisdopp)  	%原图
subplot(3,1,2)
plot(A)				%近似和
subplot(3,1,3)
plot(D)				%细节和

二维离散型哈尔小波变换

实现代码

function [ output ] = haarTranslation(img,k)
	%对img图像做k阶哈尔小波变换
	ha=[0.7071    0.7071; 0.7071   -0.7071];
	size_img = 2^int32(log2(min(size(img))));
	img = double(imresize(img,[size_img,size_img]));
	output = translation(img,k,size_img,ha);
end

function out = translation(img,k,size,ha)
    for i=1:2:size;
        for j=1:2:size;
            img(i:i+1,j:j+1)= ha*img(i:i+1,j:j+1)*ha';
        end
    end
    size_half = size/2;
    data = reshape(img,2,size_half,size);
    u = reshape(data(1,:,:),size_half,size);
    v = reshape(data(2,:,:),size_half,size);
    u = reshape(u',2,size_half,size_half);
    v = reshape(v',2,size_half,size_half);
    out =[reshape(u(1,:,:),size_half,size_half),reshape(u(2,:,:),size_half,size_half);reshape(v(1,:,:),size_half,size_half),reshape(v(2,:,:),size_half,size_half);];
    if k>1
       out(1:size_half,1:size_half) = translation(out(1:size_half,1:size_half),k-1,size_half,ha);
    end
end

调用

img = rgb2gray(imread('img.png'));
subplot(1,3,1)
imshow(img);	%原图
subplot(1,3,2)
imshow(uint8(haarTranslation(img,1)));	%一阶哈尔小波变换
subplot(1,3,3)
imshow(uint8(haarTranslation(img,2)));	%二阶哈尔小波变换

上面看不清，再补一张图