From 7d1577a6e6f621d5c44d5f7abeb11db15e4cb5fb Mon Sep 17 00:00:00 2001
From: Azalea <22280294+hykilpikonna@users.noreply.github.com>
Date: Sat, 21 Dec 2024 00:50:34 -0500
Subject: [PATCH] [+] k-means clustering and PCA

---
 server/colors.py | 131 +++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 131 insertions(+)
 create mode 100644 server/colors.py

diff --git a/server/colors.py b/server/colors.py
new file mode 100644
index 0000000..7e7947d
--- /dev/null
+++ b/server/colors.py
@@ -0,0 +1,131 @@
+import numpy as np
+from sklearn.cluster import KMeans
+from sklearn.decomposition import PCA
+from PIL import Image
+import matplotlib.pyplot as plt
+import json
+
+
+def generate_hv_colors(cluster_labels, sorted_indices):
+    """
+    Generate hColors and vColors from the clustered image.
+
+    Parameters:
+        cluster_labels (ndarray): The label of each pixel in the image corresponding to its color cluster.
+        sorted_indices (ndarray): The index mapping of the sorted cluster centers.
+
+    Returns:
+        hColors (ndarray): (n+1) x (n+1) array with the color indices of horizontal lines.
+        vColors (ndarray): (n+1) x (n+1) array with the color indices of vertical lines.
+    """
+    h, w = cluster_labels.shape
+    hColors = np.zeros((h + 1, w + 1), dtype=np.uint8)
+    vColors = np.zeros((h + 1, w + 1), dtype=np.uint8)
+
+    # Fill hColors and vColors
+    for y in range(h):
+        for x in range(w):
+            color_idx = cluster_labels[y, x]
+            color_idx = np.where(sorted_indices == color_idx)[0][0]
+            hColors[y, x] = color_idx
+            # if x >= w - 1:
+            #     hColors[y, x + 1] = color_idx
+
+            vColors[y, x] = color_idx
+            # if y >= h - 1:
+            #     vColors[y + 1, x] = color_idx
+
+    # Handle the bottom-right corner
+    # hColors[h, w] = cluster_labels[h - 1, w - 1]
+    # vColors[h, w] = cluster_labels[h - 1, w - 1]
+
+    return hColors, vColors
+
+
+def convert_colors_to_hex(cluster_centers):
+    """
+    Convert RGB colors to hex color strings.
+
+    Parameters:
+        cluster_centers (ndarray): The array of RGB colors (k, 3).
+
+    Returns:
+        list: List of hex color strings.
+    """
+    return [f"#{r:02x}{g:02x}{b:02x}" for r, g, b in cluster_centers]
+
+
+if __name__ == '__main__':
+    # Load image and convert to RGB
+    image_path = r"C:\Users\Azalea\Downloads\Untitled_Artwork.png"
+    image = Image.open(image_path).convert('RGB')
+    image_array = np.array(image)
+
+    # Reshape the image to a 2D array (num_pixels, 3)
+    height, width, _ = image_array.shape
+    pixels = image_array.reshape((-1, 3))
+
+    # Cluster the pixels using KMeans
+    n_colors = 32
+    kmeans = KMeans(n_clusters=n_colors, random_state=42)
+    kmeans.fit(pixels)
+
+    # Get the mean color of each cluster (centroids)
+    cluster_centers = kmeans.cluster_centers_.astype(int)
+
+    # Replace each pixel with the closest cluster center
+    labels = kmeans.labels_.reshape(height, width)
+    new_image_array = cluster_centers[labels].reshape(image_array.shape)
+
+    # Use PCA to sort the colors
+    pca = PCA(n_components=1)
+    color_positions = pca.fit_transform(cluster_centers)
+    sorted_indices = np.argsort(color_positions[:, 0])
+    sorted_cluster_centers = cluster_centers[sorted_indices]
+
+    # Create a color palette from the sorted cluster centers
+    palette_height = 20 * n_colors  # Each color swatch height
+    palette_width = 100  # Width of each color swatch
+    palette = np.zeros((palette_height, palette_width, 3), dtype=np.uint8)
+
+    for i, color in enumerate(sorted_cluster_centers):
+        start_y = i * 20
+        end_y = (i + 1) * 20
+        palette[start_y:end_y, :] = color
+
+    # Display the original, reduced color image, and palette
+    fig, axes = plt.subplots(1, 3, figsize=(15, 6))
+
+    axes[0].set_title('Original Image')
+    axes[0].imshow(image)
+    axes[0].axis('off')
+
+    axes[1].set_title(f'Image with {n_colors} Colors')
+    axes[1].imshow(new_image_array)
+    axes[1].axis('off')
+
+    axes[2].set_title('Sorted Color Palette (PCA)')
+    axes[2].imshow(palette)
+    axes[2].axis('off')
+
+    plt.tight_layout()
+    plt.show()
+
+    # Generate hColors and vColors
+    hColors, vColors = generate_hv_colors(labels, sorted_indices)
+
+    # Prepare output JSON structure
+    output_data = {
+        "hColors": {"__typedArray__": True, "type": "Int8Array",
+                    "data": hColors.flatten().tolist()},
+        "vColors": {"__typedArray__": True, "type": "Int8Array",
+                    "data": vColors.flatten().tolist()},
+        "colors": convert_colors_to_hex(sorted_cluster_centers)
+    }
+
+    # Save to JSON file
+    output_path = r"C:\Users\Azalea\Downloads\output.json"
+    with open(output_path, 'w') as f:
+        json.dump(output_data, f)
+
+    print(f"Saved output to {output_path}")