Picassostyle Image Maker

将一张RGB彩图处理成毕加索风格的抽象画(参数可量化), 与该过程的可视化。Process an RGB color drawing into a Picasso-style abstract painting and a gif. visualizing the process

function imgOut = picassoStyle(filename, varargin)

%picassoStyle 将一张RGB图像转换成毕加索立体主义风格的抽象画。

%picassoStyle Convert an RGB image into a Picasso-style Cubist abstract painting.

%

% 该函数通过分层处理实现艺术效果：

% This function achieves artistic effects through layered processing:

% 1. 结构抽象化：使用超像素将图像分割成几何区域。

% Structural abstraction: segment the image into geometric regions using superpixels.

% 2. 轮廓提取：从区域边界和图像主边缘提取并强化线条。

% Contour extraction: extract and enhance lines from region boundaries and main edges.

% 3. 区域着色：使用简化的调色板为区域填充平涂色块。

% Region coloring: fill regions with flat color blocks using a simplified palette.

% 4. (可选) 形式解构：对部分区域进行几何变换，模拟立体主义。

% (Optional) Form deconstruction: apply geometric transformations to regions to simulate Cubism.

%

% USAGE

% Imshow(picassoStyle(my_photo.jpg))

% imgOut = picassoStyle('my_photo.jpg');

% imgOut = picassoStyle('my_photo.jpg', 'num_superpixels', 250, 'palette_size', 12, 'line_thickness', 2);

% imgOut = picassoStyle('my_photo.jpg', 'enable_distortion', true, 'distortion_prob', 0.2);

%

% PARAMETERS

% 'num_superpixels' - 控制图像分割的块数。值越小，色块越大，画面越抽象。(默认: 300)

% Controls the number of segmented regions. Smaller values → larger blocks, more abstract. (Default: 300)

% 'palette_size' - 简化调色板中的颜色数量。值越小，颜色越抽象。(默认: 15)

% Number of colors in the simplified palette. Smaller values → more abstract colors. (Default: 15)

% 'line_thickness' - 黑色轮廓线的粗细（像素）。(默认: 1)

% Thickness of black contour lines (pixels). (Default: 1)

% 'canny_threshold' - Canny边缘检测的阈值，用于捕捉主体轮廓。(默认: 0.1)

% Threshold for Canny edge detection to capture main outlines. (Default: 0.1)

% 'enable_distortion' - 是否启用第四层“解构”效果。(默认: false)

% Enable the fourth-layer "deconstruction" effect. (Default: false)

% 'distortion_prob' - 每个区域被扭曲的概率。(默认: 0.15)

% Probability that each region will be distorted. (Default: 0.15)

% 'distortion_strength' - 扭曲的强度（旋转角度范围）。(默认: 10度)

% Distortion strength (rotation angle range). (Default: 10 degrees)

% --- 1. 参数解析与默认值设定 ---

% --- 1. Parameter parsing and default value setting ---

p = inputParser;

addRequired(p, 'filename', @(x) ischar(x) || isstring(x));

addParameter(p, 'num_superpixels', 20, @isnumeric);

addParameter(p, 'palette_size', 5, @isnumeric);

addParameter(p, 'line_thickness', 2, @isnumeric);

addParameter(p, 'canny_threshold', 0.1, @isnumeric);

addParameter(p, 'enable_distortion', true, @islogical);

addParameter(p, 'distortion_prob', 0.50, @isnumeric);

addParameter(p, 'distortion_strength', 190, @isnumeric);

parse(p, filename, varargin{:});

% 将解析的参数赋值给变量

% Assign parsed parameters to variables

num_superpixels = p.Results.num_superpixels;

palette_size = p.Results.palette_size;

line_thickness = p.Results.line_thickness;

canny_threshold = p.Results.canny_threshold;

enable_distortion = p.Results.enable_distortion;

distortion_prob = p.Results.distortion_prob;

distortion_strength = p.Results.distortion_strength;

% 读取图像

% Read image

img_orig = imread(filename);

img = im2double(img_orig); % 转换为double类型方便计算

% Convert to double type for easier computation

fprintf('开始处理图像...\n');

% Start processing image...

% --- 第一层：结构抽象化 - 图像分割与几何化 ---

% --- Layer 1: Structural abstraction - image segmentation and geometrization ---

fprintf('第一层：正在进行超像素分割...\n');

% Layer 1: Performing superpixel segmentation...

% 使用SLIC算法将图像分割成N个区域

% Use SLIC algorithm to segment image into N regions

% L是标签矩阵，每个像素的值对应其所属的区域编号

% L is a label matrix where each pixel's value corresponds to its region index

% N是实际生成的超像素数量

% N is the actual number of generated superpixels

[L, N] = superpixels(img, num_superpixels, 'Compactness', 20);

% --- 第二层：轮廓提取与强化 - 勾勒关键线条 ---

% --- Layer 2: Contour extraction and enhancement - outline key lines ---

fprintf('第二层：正在提取和强化轮廓线...\n');

% Layer 2: Extracting and enhancing contour lines...

% 1. 提取超像素区域之间的边界

% 1. Extract boundaries between superpixel regions

superpixel_boundaries = boundarymask(L);

% 2. 提取原图的主体轮廓

% 2. Extract main outlines from the original image

gray_img = rgb2gray(img);

canny_edges = edge(gray_img, 'canny', canny_threshold);

% 3. 合并两种边界并加粗，形成最终的线条层

% 3. Combine both boundaries and thicken them to form the final line layer

combined_lines = superpixel_boundaries | canny_edges;

se = strel('disk', line_thickness);

line_layer = imdilate(combined_lines, se);

% 将线条层反转，用于后续叠加（黑色线条区域为0）

% Invert line layer for later overlay (black line regions as 0)

line_mask = ~line_layer;

% --- 第三层：区域着色与色彩简化 - 大胆的色块填充 ---

% --- Layer 3: Region coloring and palette simplification - bold color filling ---

fprintf('第三层：正在简化调色板并填充色块...\n');

% Layer 3: Simplifying palette and filling regions with colors...

% 1. 创建简化调色板

% 1. Create simplified palette

% 使用rgb2ind从原图中提取有限数量的主导色

% Use rgb2ind to extract a limited set of dominant colors from the original image

[~, simple_map] = rgb2ind(img_orig, palette_size, 'nodither');

% 2. 为每个超像素区域确定一个简化颜色并填充

% 2. Assign each superpixel region a simplified color and fill it

% 初始化一个空白的颜色填充层

% Initialize an empty color fill layer

color_fill_layer = zeros(size(img));

% label2idx可以高效地获取每个区域的像素索引

% label2idx efficiently retrieves pixel indices for each region

idx = label2idx(L);

for i = 1:N

region_idx = idx{i};

% 计算该区域在原图中的平均颜色

% Compute mean color of this region in the original image

mean_color = mean(reshape(img(repmat(L==i, [1,1,3])), [], 3), 1);

% 从简化调色板中找到与该平均色最接近的颜色

% Find the closest color in the simplified palette

[~, closest_color_idx] = min(sum((simple_map - mean_color).^2, 2));

fill_color = simple_map(closest_color_idx, :);

% 在颜色填充层中，将该区域所有像素设置为找到的颜色

% Assign this color to all pixels of the region in the color fill layer

% 需要分别对R, G, B三个通道进行操作

% Must operate separately on R, G, B channels

for channel = 1:3

channel_plane = color_fill_layer(:,:,channel);

channel_plane(region_idx) = fill_color(channel);

color_fill_layer(:,:,channel) = channel_plane;

end

end

% --- 第四层（可选）：形式解构与重组 ---

% --- Layer 4 (optional): Form deconstruction and reconstruction ---

output_image = color_fill_layer; % 默认输出为第三层的结果

% Default output is result from Layer 3

if enable_distortion

fprintf('第四层：正在进行形式解构（扭曲）...\n');

% Layer 4: Performing form deconstruction (distortion)...

distorted_layer = output_image;

for i = 1:N

% 以一定概率决定是否扭曲当前区域

% With certain probability, decide whether to distort current region

if rand() < distortion_prob

% 创建一个随机的微小仿射变换（这里是旋转）

% Create a random small affine transformation (here, rotation)

angle = (rand - 0.5) * 2 * distortion_strength;

tform = affine2d([cosd(angle) -sind(angle) 0; sind(angle) cosd(angle) 0; 0 0 1]);

% 获取当前区域的掩码和颜色

% Get mask and color of the current region

region_mask = (L == i);

% 从已经上色的图中获取颜色

% Get colors from already filled image

color_plane = output_image .* double(region_mask);

% 对该区域进行扭曲

% Distort this region

% 'OutputView'确保输出与原图大小一致

% 'OutputView' ensures output matches original size

R = imref2d(size(img));

warped_plane = imwarp(color_plane, tform, 'OutputView', R);

warped_mask = imwarp(region_mask, tform, 'OutputView', R);

% 将扭曲后的区域“贴”回画布

% Paste the distorted region back onto the canvas

distorted_layer(warped_mask) = warped_plane(warped_mask);

end

end

output_image = distorted_layer;

end

% --- 最终合成 ---

% --- Final composition ---

fprintf('正在合成最终图像...\n');

% Composing final image...

% 将黑色线条叠加在彩色图层上

% Overlay black lines on the colored layer

% line_mask是一个逻辑矩阵，需要扩展到3个通道

% line_mask is a logical matrix, needs expansion to 3 channels

final_image = output_image .* double(repmat(line_mask, [1,1,3]));

% 转换回uint8格式

% Convert back to uint8 format

imgOut = im2uint8(final_image);

fprintf('处理完成！\n');

% Processing complete!

end

%%_______GIF. Making Code(should be saved as another document to run)_________

%% makeGif.m

% 基于 picassoStyle.m 生成动图

% MATLAB R2023a

clc; clear; close all;

% 输入图像

filename = "My_photo.jpg"; % 可以换成自己的图像

% GIF 文件名

gif_filename = 'picasso_animation.gif';

% 区块数量范围：200 → 10，步长 -10

block_nums = 200:-10:10;

% Canny 阈值范围：0.1 → 0.9

canny_vals = linspace(0.1, 0.9, numel(block_nums));

% 动画时长和帧率

nFrames = numel(block_nums);

totalDuration = 5; % 秒

frameDelay = totalDuration / nFrames; % 每帧延时

for i = 1:nFrames

numBlocks = block_nums(i);

cannyTh = canny_vals(i);

% 调用 picassoStyle

imgOut = picassoStyle(filename, ...

'num_superpixels', numBlocks, ...

'palette_size', 20, ...

'line_thickness', 2, ...

'canny_threshold', cannyTh, ...

'enable_distortion', false);

% 显示当前帧

imshow(imgOut);

title(sprintf('区块数: %d, Canny阈值: %.2f', numBlocks, cannyTh), 'FontSize', 14);

drawnow;

% 捕获并写入 GIF

frame = getframe(gca);

[imind, cm] = rgb2ind(frame.cdata, 256);

if i == 1

imwrite(imind, cm, gif_filename, 'gif', 'Loopcount', inf, 'DelayTime', frameDelay);

else

imwrite(imind, cm, gif_filename, 'gif', 'WriteMode', 'append', 'DelayTime', frameDelay);

end

end