Morandi Palette: A palette with 18 colors inspired by Giorgio Morandi

morandi.py

import os
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt
from collections import defaultdict
from gurobipy import Model, GRB
from PIL import Image

colors = [
    '#686789', '#B77F70', '#E5E2B9', '#BEB1A8', '#A79A89', '#8A95A9', 
    '#ECCED0', '#7D7465', '#E8D3C0', '#7A8A71', '#789798', '#B57C82', 
    '#9FABB9', '#B0B1B6', '#99857E', '#88878D', '#91A0A5', '#9AA690'
    ]

def hex_to_rgb(value):
    """Convert a hex color to an RGB tuple."""
    value = value.lstrip("#")
    return tuple(int(value[i : i + 2], 16) for i in (0, 2, 4))


def rgb_to_hex(rgb):
    return "#{:02x}{:02x}{:02x}".format(rgb[0], rgb[1], rgb[2])


def euclidean_distance(color1, color2):
    """Calculate the Euclidean distance between two RGB colors."""
    return int(sum((c1 - c2) ** 2 for c1, c2 in zip(color1, color2)))


def find_next_color(palette, remaining_colors):
    """Find the color from remaining_colors that maximizes the average distance to the current palette."""
    max_avg_distance = 0
    next_color = None
    for color in remaining_colors:
        print(color, palette)
        avg_distance = sum(
            euclidean_distance(hex_to_rgb(color), hex_to_rgb(p)) for p in palette
        ) / len(palette)
        if avg_distance > max_avg_distance:
            max_avg_distance = avg_distance
            next_color = color
    return next_color


def sort_colors(colors):
    """This is only to demonstrate the algorithm to get the ordered colors."""
    current_palette = colors[:6]

    remaining_colors = [color for color in colors if color not in current_palette]
    for _ in range(len(remaining_colors)):
        next_color = find_next_color(current_palette, remaining_colors)
        current_palette.append(next_color)
        remaining_colors.remove(next_color)
    return current_palette


def get_colors(dir: str, num: int = 18, delta: int = 1500):
    """Get a list of colors from the Morandi palette."""
    if dir is None:
        return colors[:num]

    color_counts = defaultdict(int)
    for filename in os.listdir(dir):
        if filename.endswith(".jpg") or filename.endswith(".png"):
            filepath = os.path.join(dir, filename)
            print(f"Processing {filepath}")
            with Image.open(filepath) as img:
                img = img.resize((100, 100))
                img = img.convert("RGB")
                color_sig = img.getcolors(img.size[0] * img.size[1])
                for count, color in color_sig:
                    matched = False
                    for existing_color in color_counts:
                        distance = euclidean_distance(existing_color, color)
                        if distance < delta:
                            color_counts[existing_color] += count
                            matched = True
                            break
                    if not matched:
                        color_counts[color] = count

    top_colors = sorted(color_counts.items(), key=lambda x: x[1], reverse=True)[:num]
    top_colors_hex = [(rgb_to_hex(color), count) for color, count in top_colors]
    return [color for color, _ in top_colors_hex]


def group_colors(colors, num_per_group):
    """Group colors into groups of fixed size such that the sum of the distances between each pair of colors in the same group is maximized."""

    # Create a new model
    m = Model("color_grouping")

    # Create variables
    x = [
        [m.addVar(vtype=GRB.BINARY, name=f"{i}-{j}") for j in range(len(num_per_group))]
        for i in range(len(colors))
    ]

    # Set objective
    objective = 0
    for i in range(len(colors)):
        for j in range(i + 1, len(colors)):
            for k in range(len(num_per_group)):
                objective += (
                    x[i][k]
                    * x[j][k]
                    * euclidean_distance(hex_to_rgb(colors[i]), hex_to_rgb(colors[j]))
                )
    m.setObjective(objective, GRB.MAXIMIZE)
    m.setParam("TimeLimit", 5 * 60)

    # Add constraints
    for i in range(len(colors)):
        m.addConstr(sum(x[i]) <= 1)  # each color can only be at most in one group

    for j in range(len(num_per_group)):
        m.addConstr(
            sum(x[i][j] for i in range(len(colors))) == num_per_group[j]
        )  # each group has a fixed number of colors

    # Optimize model
    m.optimize()

    # Check if a solution exists
    if m.status in [GRB.OPTIMAL, GRB.SUBOPTIMAL, GRB.TIME_LIMIT]:
        (
            print("Warning: suboptimal solution found")
            if m.status == GRB.SUBOPTIMAL
            else None
        )
        result = [[] for _ in range(len(num_per_group))]
        for i in range(len(colors)):
            for k in range(len(num_per_group)):
                if x[i][k].x == 1:
                    result[k].append(colors[i])
        return result
    else:
        raise Exception("No solution found")


if __name__ == "__main__":
    ## Example 1: getting ordered K-colors
    colors = sort_colors(get_colors(None, 18))
    sns.palplot(colors)
    plt.title("Giorgio Morandi's palette")
    plt.show()

    ## Example 2: getting grouped K-colors
    groups = group_colors(colors, [2, 2, 2, 1, 2, 2, 3, 2, 1])

    print("Grouped colors:", groups)

    count, n = 0, len(groups)
    fig, axes = plt.subplots(1, n, figsize=(15, 2))
    for ax, group in zip(axes, groups):
        ax.imshow(
            [[np.array(hex_to_rgb(color)) / 255.0 for color in group]], aspect="auto"
        )
        ax.set_title(f"Group-{count}")
        ax.axis("off")
        count += 1
    plt.show()