Skip to content
TheChatArchiveThe Chat Archive
gpt-3.5-turbo-0301· OpenAIAi2 WildChat-4.8M (AI2)

给我CLAHE的算法代码

  1. User

    给我CLAHE的算法代码

  2. Assistant

    以下是Python实现的CLAHE算法代码:

    import cv2
    import numpy as np
    
    def clahe(img, clip_limit=2.0, grid_size=(8, 8)):
        h, w = img.shape[:2]
        grid_h, grid_w = grid_size
        
        # 计算每个网格的直方图
        hist_grid = np.zeros((grid_h, grid_w, 256), dtype=np.uint32)
        for i in range(grid_h):
            for j in range(grid_w):
                y1 = i * h // grid_h
                y2 = (i + 1) * h // grid_h
                x1 = j * w // grid_w
                x2 = (j + 1) * w // grid_w
                hist, _ = np.histogram(img[y1:y2, x1:x2].flat, bins=256, range=(0, 255))
                hist_grid[i, j] = hist
        
        # 计算每个网格的累计直方图和每个像素所在的网格
        cdf_grid = np.zeros((grid_h, grid_w, 256), dtype=np.uint32)
        grid_idx = np.zeros((h, w), dtype=np.uint32)
        for i in range(grid_h):
            for j in range(grid_w):
                cdf = np.cumsum(hist_grid[i, j])
                cdf = np.uint32(255 * cdf / cdf[-1])
                cdf_grid[i, j] = cdf
                y1 = i * h // grid_h
                y2 = (i + 1) * h // grid_h
                x1 = j * w // grid_w
                x2 = (j + 1) * w // grid_w
                grid_idx[y1:y2, x1:x2] = i * grid_w + j
        
        # 对每个像素进行CLAHE处理
        clip = np.uint32(clip_limit * grid_h * grid_w / 256)
        for i in range(h):
            for j in range(w):
                gx, gy = grid_idx[i, j] // grid_w, grid_idx[i, j] % grid_w
                cdf = cdf_grid[gx, gy]
                val = img[i, j]
                if cdf[val] <= clip:
                    img[i, j] = cdf[clip]
                else:
                    img[i, j] = cdf[val]
        
        return img
    

    使用方法:调用clahe函数并传入需要处理的图像即可。clip_limit和grid_size分别是对比度限制因子和网格大小,可以根据需要进行调整。

  3. User

    转为matlab

  4. Assistant

    以下是MATLAB实现的CLAHE算法代码:

    function img_out = clahe(img_in, clip_limit, grid_size)
    
    % 获取图像尺寸和网格尺寸
    [h, w] = size(img_in);
    grid_h = grid_size(1);
    grid_w = grid_size(2);
    
    % 计算每个网格的直方图
    hist_grid = zeros(grid_h, grid_w, 256);
    for i = 1:grid_h
        for j = 1:grid_w
            y1 = floor((i-1) * h / grid_h) + 1;
            y2 = floor(i * h / grid_h);
            x1 = floor((j-1) * w / grid_w) + 1;
            x2 = floor(j * w / grid_w);
            hist = histcounts(img_in(y1:y2, x1:x2), 256, 'BinLimits', [0 255]);
            hist_grid(i, j, :) = hist;
        end
    end
    
    % 计算每个网格的累计直方图和每个像素所在的网格
    cdf_grid = zeros(grid_h, grid_w, 256);
    grid_idx = zeros(h, w);
    for i = 1:grid_h
        for j = 1:grid_w
            cdf = cumsum(hist_grid(i, j, :));
            cdf = uint32(255 * cdf / cdf(end));
            cdf_grid(i, j, :) = cdf;
            y1 = floor((i-1) * h / grid_h) + 1;
            y2 = floor(i * h / grid_h);
            x1 = floor((j-1) * w / grid_w) + 1;
            x2 = floor(j * w / grid_w);
            grid_idx(y1:y2, x1:x2) = (i-1) * grid_w + j;
        end
    end
    
    % 对每个像素进行CLAHE处理
    clip = uint32(clip_limit * grid_h * grid_w / 256);
    img_out = img_in;
    for i = 1:h
        for j = 1:w
            idx = grid_idx(i, j);
            gx = floor((idx-1) / grid_w) + 1;
            gy = mod(idx-1, grid_w) + 1;
            cdf = cdf_grid(gx, gy, :);
            val = img_in(i, j);
            if cdf(val+1) <= clip
                img_out(i, j) = cdf(clip+1);
            else
                img_out(i, j) = cdf(val+1);
            end
        end
    end
    
    end
    

    使用方法:调用clahe函数并传入需要处理的图像、对比度限制因子和网格大小即可。