簡介
在螢幕上看起來不錯的設計在列印上可能並不好看。本文處理列印製作流程：標準 RGB → CMYK 轉換，確保最低 DPI 為 300，以及使用 Real-ESRGAN 的 AI 高檔低解析度設計。
1. 印刷生產流程
User Design (RGB, 72-150 DPI)
│
▼
┌──────────────────────┐
│ 1. Quality Assessment │ ← Check DPI, resolution, artifacts
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ 2. AI Super Resolution│ ← Real-ESRGAN upscale (if needed)
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ 3. Color Conversion │ ← RGB → CMYK với ICC profile
└──────────┬───────────┘
│
▼
┌──────────────────────┐
│ 4. Print File Export │ ← TIFF/PDF, 300+ DPI, CMYK
└──────────────────────┘
2. 品質評估
from PIL import Image
import numpy as np
class PrintQualityAssessor:
"""Đánh giá chất lượng design cho print"""
# DPI thresholds cho các phương pháp in
MIN_DPI = {
"dtg": 300, # Direct-to-Garment
"screen_print": 300, # Screen Printing
"sublimation": 150, # Sublimation
"vinyl": 72, # Vinyl Cut
}
def assess(
self,
image: Image.Image,
target_size_cm: tuple[float, float],
print_method: str = "dtg",
) -> PrintQualityReport:
width_px, height_px = image.size
target_w_cm, target_h_cm = target_size_cm
# Calculate actual DPI at target size
dpi_w = width_px / (target_w_cm / 2.54)
dpi_h = height_px / (target_h_cm / 2.54)
actual_dpi = min(dpi_w, dpi_h)
required_dpi = self.MIN_DPI[print_method]
# Calculate upscale needed
if actual_dpi < required_dpi:
scale_factor = required_dpi / actual_dpi
needs_upscale = True
else:
scale_factor = 1.0
needs_upscale = False
# Check for artifacts
artifacts = self._detect_artifacts(image)
return PrintQualityReport(
actual_dpi=actual_dpi,
required_dpi=required_dpi,
needs_upscale=needs_upscale,
scale_factor=scale_factor,
artifacts=artifacts,
is_print_ready=actual_dpi >= required_dpi and not artifacts,
)
def _detect_artifacts(
self, image: Image.Image
) -> list[str]:
"""Detect common image artifacts"""
issues = []
img_array = np.array(image)
# Check JPEG artifacts (blockiness)
if self._detect_blockiness(img_array) > 0.3:
issues.append("jpeg_artifacts")
# Check if too small
if min(image.size) < 500:
issues.append("very_low_resolution")
# Check transparency issues
if image.mode == "RGBA":
alpha = np.array(image.split()[3])
if np.any((alpha > 0) & (alpha < 255)):
issues.append("semi_transparent_pixels")
return issues
3. AI超解析度（Real-ESRGAN）
import torch
from basicsr.archs.rrdbnet_arch import RRDBNet
from realesrgan import RealESRGANer
class AIUpscaler:
"""AI-powered image upscaling cho print"""
def __init__(self):
# RealESRGAN x4 model
model = RRDBNet(
num_in_ch=3, num_out_ch=3,
num_feat=64, num_block=23,
num_grow_ch=32, scale=4,
)
self.upsampler = RealESRGANer(
scale=4,
model_path="models/RealESRGAN_x4plus.pth",
model=model,
tile=400, # Process in tiles (GPU memory)
tile_pad=10,
pre_pad=0,
half=True, # FP16 for speed
)
def upscale(
self,
image: Image.Image,
target_scale: float,
) -> Image.Image:
"""
Upscale image by target_scale factor
Real-ESRGAN supports 4x natively.
For other scales, upscale 4x then resize.
"""
img_array = np.array(image.convert("RGB"))
# Upscale 4x
output, _ = self.upsampler.enhance(
img_array, outscale=4
)
result = Image.fromarray(output)
# If target < 4x, downscale to exact size
if target_scale < 4:
new_size = (
int(image.width * target_scale),
int(image.height * target_scale),
)
result = result.resize(
new_size, Image.Resampling.LANCZOS
)
return result
def upscale_for_print(
self,
image: Image.Image,
target_size_cm: tuple[float, float],
target_dpi: int = 300,
) -> Image.Image:
"""Upscale to exact print size + DPI"""
target_w_px = int(target_size_cm[0] / 2.54 * target_dpi)
target_h_px = int(target_size_cm[1] / 2.54 * target_dpi)
# Calculate needed scale
scale_w = target_w_px / image.width
scale_h = target_h_px / image.height
scale = max(scale_w, scale_h)
if scale <= 1.0:
# Already big enough, just resize
return image.resize(
(target_w_px, target_h_px),
Image.Resampling.LANCZOS,
)
# Multi-pass upscale if scale > 4x
current = image
while scale > 1.0:
step_scale = min(scale, 4.0)
current = self.upscale(current, step_scale)
scale /= 4.0
# Final resize to exact dimensions
return current.resize(
(target_w_px, target_h_px),
Image.Resampling.LANCZOS,
)
4. RGB → CMYK 轉換
from PIL import ImageCms
class ColorConverter:
"""Convert RGB → CMYK với ICC profile"""
def __init__(self):
# ICC profiles
self.srgb_profile = ImageCms.createProfile("sRGB")
self.cmyk_profile = ImageCms.getOpenProfile(
"profiles/USWebCoatedSWOP.icc"
)
def rgb_to_cmyk(
self,
image: Image.Image,
rendering_intent: int = ImageCms.Intent.RELATIVE_COLORIMETRIC,
) -> Image.Image:
"""
Convert RGB → CMYK sử dụng ICC profiles
Rendering Intents:
- PERCEPTUAL (0): Giữ overall look, tốt cho photos
- RELATIVE_COLORIMETRIC (1): Accurate colors, default
- SATURATION (2): Vivid colors, tốt cho graphics
- ABSOLUTE_COLORIMETRIC (3): Exact match
"""
# Ensure RGB input
if image.mode == "RGBA":
# Flatten alpha onto white background
background = Image.new("RGB", image.size, (255, 255, 255))
background.paste(image, mask=image.split()[3])
image = background
elif image.mode != "RGB":
image = image.convert("RGB")
# Create transform
transform = ImageCms.buildTransformFromOpenProfiles(
self.srgb_profile,
self.cmyk_profile,
"RGB",
"CMYK",
renderingIntent=rendering_intent,
)
# Apply transform
cmyk_image = ImageCms.applyTransform(image, transform)
return cmyk_image
def check_gamut(
self, image: Image.Image
) -> GamutReport:
"""Check which colors are out of CMYK gamut"""
img_array = np.array(image.convert("RGB"))
out_of_gamut = 0
total = img_array.shape[0] * img_array.shape[1]
for y in range(0, img_array.shape[0], 10):
for x in range(0, img_array.shape[1], 10):
r, g, b = img_array[y, x]
if self._is_out_of_gamut(r, g, b):
out_of_gamut += 1
return GamutReport(
total_pixels_sampled=total // 100,
out_of_gamut_pixels=out_of_gamut,
percentage=out_of_gamut / (total // 100) * 100,
)
5. 列印文件匯出管道
class PrintFileExporter:
"""Export print-ready files"""
async def export(
self,
design: Image.Image,
product_type: str,
size: str,
print_method: str = "dtg",
) -> PrintFile:
# 1. Get print specs
specs = self._get_print_specs(product_type, size)
# 2. Assess quality
quality = assessor.assess(
design, specs.print_area_cm, print_method
)
# 3. Upscale if needed
if quality.needs_upscale:
design = upscaler.upscale_for_print(
design, specs.print_area_cm, specs.dpi
)
# 4. Convert color space
cmyk_design = color_converter.rgb_to_cmyk(design)
# 5. Export TIFF
output_path = f"print_files/{uuid4()}.tiff"
cmyk_design.save(
output_path,
format="TIFF",
dpi=(specs.dpi, specs.dpi),
compression="lzw",
)
return PrintFile(
path=output_path,
dpi=specs.dpi,
color_space="CMYK",
size_cm=specs.print_area_cm,
)
總結
列印檔案優化：
- 品質評估 — DPI 檢查、偽影檢測
- Real-ESRGAN — AI 高檔 4x，多通道 > 4x
- 顏色轉換 — ICC配置檔RGB→CMYK，色域檢查
- 列印匯出 — TIFF/PDF、300 DPI、LZW 壓縮
下一篇文章：人工智慧自動標記 — 自動標記和分類設計。