图谱融合有哪些方法？并给出matlab代码

图谱融合是将不同模态的图像融合成一幅图像，以提高图像的质量和信息量。常用的图谱融合方法有以下几种：

1. 像素级融合（Pixel-level Fusion）

该方法是将不同模态的图像像素级别进行融合。常见的像素级融合方法有加权平均法、最大值法、最小值法等。

Matlab代码示例：

% 加权平均法
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
alpha = 0.5; % 权重
img3 = alpha * img1 + (1 - alpha) * img2;
imshow(img3);

% 最大值法
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
img3 = max(img1, img2);
imshow(img3);

% 最小值法
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
img3 = min(img1, img2);
imshow(img3);

2. 特征级融合（Feature-level Fusion）

该方法是将不同模态的图像的特征进行融合，常见的特征包括纹理特征、形状特征、颜色特征等。常用的特征级融合方法有主成分分析法（PCA）、小波变换法等。

Matlab代码示例：

% PCA
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
img1 = double(rgb2gray(img1));
img2 = double(rgb2gray(img2));
[coeff1, ~] = pca(img1);
[coeff2, ~] = pca(img2);
alpha = 0.5; % 权重
coeff3 = alpha * coeff1 + (1 - alpha) * coeff2;
img3 = img1 * coeff3 * coeff3' + img2 * (eye(size(coeff3, 1)) - coeff3 * coeff3');
imshow(uint8(img3));

% 小波变换
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
[cA1, cH1, cV1, cD1] = dwt2(rgb2gray(img1), 'haar');
[cA2, cH2, cV2, cD2] = dwt2(rgb2gray(img2), 'haar');
alpha = 0.5; % 权重
cA3 = alpha * cA1 + (1 - alpha) * cA2;
cH3 = alpha * cH1 + (1 - alpha) * cH2;
cV3 = alpha * cV1 + (1 - alpha) * cV2;
cD3 = alpha * cD1 + (1 - alpha) * cD2;
img3 = idwt2(cA3, cH3, cV3, cD3, 'haar');
imshow(uint8(img3));

3. 决策级融合（Decision-level Fusion）

该方法是将不同模态的图像的决策进行融合，常用的决策包括分类结果、检测结果、分割结果等。常见的决策级融合方法有投票法、加权投票法、D-S证据理论等。

Matlab代码示例：

% 投票法
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
img1 = double(rgb2gray(img1));
img2 = double(rgb2gray(img2));
img1(img1 < 128) = 0; img1(img1 >= 128) = 1; % 二值化
img2(img2 < 128) = 0; img2(img2 >= 128) = 1;
img3 = img1 + img2;
img3(img3 < 2) = 0; img3(img3 >= 2) = 1; % 投票
imshow(uint8(img3 * 255));

% 加权投票法
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
img1 = double(rgb2gray(img1));
img2 = double(rgb2gray(img2));
img1(img1 < 128) = 0; img1(img1 >= 128) = 1; % 二值化
img2(img2 < 128) = 0; img2(img2 >= 128) = 1;
alpha = 0.5; % 权重
img3 = alpha * img1 + (1 - alpha) * img2; % 加权投票
img3(img3 < 0.5) = 0; img3(img3 >= 0.5) = 1;
imshow(uint8(img3 * 255));

% D-S证据理论
img1 = imread('image1.jpg');
img2 = imread('image2.jpg');
img1 = double(rgb2gray(img1));
img2 = double(rgb2gray(img2));
img1(img1 < 128) = 0; img1(img1 >= 128) = 1; % 二值化
img2(img2 < 128) = 0; img2(img2 >= 128) = 1;
P1 = img1(:); P2 = img2(:); % 转换为向量
P = [P1, P2];
[m, n] = size(P);
D = zeros(m, m); % 计算D-S证据
for i = 1:m
    for j = 1:m
        if i == j
            continue;
        end
        D(i, j) = sum(P(i, :) & ~P(j, :)) / sum(P(i, :) | P(j, :));
    end
end
W = zeros(1, m); % 计算权重
for i = 1:m
    W(i) = sum(D(i, :)) / (m - 1);
end
alpha = W(1) / sum(W); % 计算权重
img3 = alpha * img1 + (1 - alpha) * img2; % 加权投票
img3(img3 < 0.5) = 0; img3(img3 >= 0.5) = 1;
imshow(uint8(img3 * 255));