高斯混合模型实现图像分割matlab代码

以下是一个简单的高斯混合模型实现图像分割的Matlab代码示例：

image = imread('image.jpg');
img_double = im2double(image);
[m, n, ~] = size(img_double);
num_clusters = 5; % 设置聚类数目

% 初始化高斯混合模型参数
mu = rand(num_clusters, 3);
sigma = repmat(eye(3), [1, 1, num_clusters]);
w = ones(1, num_clusters) / num_clusters;

% 迭代
max_iter = 100;
for iter = 1:max_iter
    % E步：计算每个像素属于每个高斯分布的概率
    prob = zeros(m*n, num_clusters);
    for i = 1:num_clusters
        prob(:, i) = w(i) * mvnpdf(reshape(img_double, [m*n, 3]), mu(i, :), sigma(:, :, i));
    end
    prob_sum = sum(prob, 2);
    gamma = prob ./ repmat(prob_sum, [1, num_clusters]);

    % M步：更新高斯混合模型参数
    total_gamma = sum(gamma, 1);
    for i = 1:num_clusters
        mu(i, :) = (1 / total_gamma(i)) * sum(gamma(:, i) .* img_double, 1);
        sigma(:, :, i) = (1 / total_gamma(i)) * (gamma(:, i)' * (img_double - mu(i, :)) .* (img_double - mu(i, :)));
        w(i) = total_gamma(i) / (m * n);
    end
end

% 计算每个像素属于哪个高斯分布
prob = zeros(m*n, num_clusters);
for i = 1:num_clusters
    prob(:, i) = w(i) * mvnpdf(reshape(img_double, [m*n, 3]), mu(i, :), sigma(:, :, i));
end
[~, labels] = max(prob, [], 2);
labels = reshape(labels, [m, n]);

% 显示分割结果
figure;
imshow(labels, []);

以上代码中，`image` 是输入的图像，`num_clusters` 是指定的高斯混合模型中高斯分布的数量。初始化高斯混合模型参数后，使用 EM 算法迭代估计参数，在每次迭代中，先计算每个像素属于每个高斯分布的概率，然后根据这些概率更新高斯混合模型参数。最后，根据更新后的参数计算每个像素属于哪个高斯分布并进行分割。