#!/usr/bin/env python3 """Create transparent-background product image copies. The script never overwrites source files. It removes only near-white pixels that are connected to the image edges, plus desaturated studio-background shadows. That protects white labels inside jars because they are not connected to the outer image border. """ from __future__ import annotations import argparse import json import re from collections import deque from pathlib import Path from typing import Iterable from PIL import Image, ImageFilter THUMB_RE = re.compile(r"-\d+x\d+\.(jpe?g|png|webp)$", re.IGNORECASE) try: RESAMPLE_LANCZOS = Image.Resampling.LANCZOS RESAMPLE_BILINEAR = Image.Resampling.BILINEAR except AttributeError: RESAMPLE_LANCZOS = Image.LANCZOS RESAMPLE_BILINEAR = Image.BILINEAR def iter_images(source: Path) -> Iterable[Path]: for path in sorted(source.rglob("*")): if not path.is_file(): continue if path.suffix.lower() not in {".jpg", ".jpeg", ".png", ".webp"}: continue if "woocommerce-placeholder" in path.name: continue if THUMB_RE.search(path.name): continue yield path def is_background( pixel: tuple[int, int, int, int], threshold: int, tolerance: int, shadow_luma: int, shadow_chroma: int, ) -> bool: r, g, b, a = pixel if a == 0: return True chroma = max(r, g, b) - min(r, g, b) luma = int((0.2126 * r) + (0.7152 * g) + (0.0722 * b)) if r >= threshold and g >= threshold and b >= threshold and chroma <= tolerance: return True # Product photos often have grey/off-white paper shadows connected to the # edge. Remove those only when they are low-chroma, so fruit, cloth, labels, # caps, and dark jar contents are preserved. return luma >= shadow_luma and chroma <= shadow_chroma def edge_connected_mask( image: Image.Image, threshold: int, tolerance: int, shadow_luma: int, shadow_chroma: int, protect_radius: int, ) -> Image.Image: rgba = image.convert("RGBA") width, height = rgba.size total = width * height candidates = bytearray(total) for idx, pixel in enumerate(rgba.getdata()): if is_background(pixel, threshold, tolerance, shadow_luma, shadow_chroma): candidates[idx] = 1 protected = bytearray(total) if protect_radius > 0: foreground = Image.frombytes( "L", (width, height), bytes(0 if candidate else 255 for candidate in candidates), ) size = max(3, (protect_radius * 2) + 1) protected = bytearray(foreground.filter(ImageFilter.MaxFilter(size)).tobytes()) mask_bytes = bytearray(total) queue: deque[int] = deque() def push(idx: int) -> None: if mask_bytes[idx] or not candidates[idx] or protected[idx]: return mask_bytes[idx] = 255 queue.append(idx) for x in range(width): push(x) push((height - 1) * width + x) for y in range(height): push(y * width) push(y * width + width - 1) while queue: idx = queue.popleft() x = idx % width if x > 0: push(idx - 1) if x < width - 1: push(idx + 1) if idx >= width: push(idx - width) if idx < total - width: push(idx + width) return Image.frombytes("L", (width, height), bytes(mask_bytes)) def process_image( source: Path, output: Path, threshold: int, tolerance: int, shadow_luma: int, shadow_chroma: int, feather: float, expand: int, protect_radius: int, max_edge: int, ) -> dict: image = Image.open(source).convert("RGBA") work = image.copy() if max_edge and max(image.size) > max_edge: work.thumbnail((max_edge, max_edge), RESAMPLE_LANCZOS) mask = edge_connected_mask(work, threshold, tolerance, shadow_luma, shadow_chroma, protect_radius) if expand > 0: size = max(3, (expand * 2) + 1) mask = mask.filter(ImageFilter.MaxFilter(size)) if feather > 0: mask = mask.filter(ImageFilter.GaussianBlur(feather)) if mask.size != image.size: mask = mask.resize(image.size, RESAMPLE_BILINEAR) alpha = image.getchannel("A") alpha = Image.composite(Image.new("L", image.size, 0), alpha, mask) result = image.copy() result.putalpha(alpha) output.parent.mkdir(parents=True, exist_ok=True) result.save(output, "PNG", optimize=True) removed_pixels = sum(1 for value in mask.getdata() if value > 0) total_pixels = image.size[0] * image.size[1] return { "source": str(source), "output": str(output), "removed_ratio": round(removed_pixels / total_pixels, 4) if total_pixels else 0, "width": image.size[0], "height": image.size[1], } def main() -> int: parser = argparse.ArgumentParser() parser.add_argument("--source", default="storage/app/public/products") parser.add_argument("--output", default="storage/app/public/products-cutout") parser.add_argument("--manifest", default="storage/app/product-cutouts-manifest.json") parser.add_argument("--limit", type=int, default=0) parser.add_argument("--image", action="append", default=[]) parser.add_argument("--list-file", default="") parser.add_argument("--max-edge", type=int, default=1600) parser.add_argument("--threshold", type=int, default=238) parser.add_argument("--tolerance", type=int, default=18) parser.add_argument("--shadow-luma", type=int, default=184) parser.add_argument("--shadow-chroma", type=int, default=34) parser.add_argument("--feather", type=float, default=0.8) parser.add_argument("--expand", type=int, default=1) parser.add_argument("--protect-radius", type=int, default=10) parser.add_argument("--dry-run", action="store_true") parser.add_argument("--skip-existing", action="store_true") args = parser.parse_args() source_root = Path(args.source).resolve() output_root = Path(args.output) manifest_path = Path(args.manifest) records = [] if args.image: images = [Path(path) if Path(path).is_absolute() else Path.cwd() / path for path in args.image] elif args.list_file: list_root = Path.cwd() images = [] for line in Path(args.list_file).read_text(encoding="utf-8").splitlines(): line = line.strip() if not line: continue path = Path(line) images.append(path if path.is_absolute() else list_root / path) else: images = list(iter_images(source_root)) if args.limit: images = images[: args.limit] for source in images: try: relative = source.resolve().relative_to(source_root) except ValueError: relative = Path(source.name) output = output_root / relative.with_suffix(".png") if source.suffix.lower() not in {".jpg", ".jpeg", ".png", ".webp"} or not source.exists(): records.append({"source": str(source), "output": str(output), "skipped": "missing_or_unsupported"}) continue if args.skip_existing and output.exists(): records.append({"source": str(source), "output": str(output), "skipped": "exists"}) continue if args.dry_run: records.append({"source": str(source), "output": str(output), "dry_run": True}) continue records.append(process_image( source, output, args.threshold, args.tolerance, args.shadow_luma, args.shadow_chroma, args.feather, args.expand, args.protect_radius, args.max_edge, )) manifest_path.parent.mkdir(parents=True, exist_ok=True) manifest_path.write_text(json.dumps({ "source_root": str(source_root), "output_root": str(output_root), "count": len(records), "records": records, }, ensure_ascii=False, indent=2), encoding="utf-8") print(f"{'Would process' if args.dry_run else 'Processed'} {len(records)} images") print(f"Manifest: {manifest_path}") return 0 if __name__ == "__main__": raise SystemExit(main())