如何使用OpenCV（Java）为文档图像添加色彩效果

Hey there! Since you're already comfortable with those blur and edge detection filters in OpenCV, adding color effects to your document scans should be totally manageable. Let's break down some practical, document-friendly color adjustments you can implement right away:

Documents often look best in high-contrast black and white, which cuts down on noise and improves readability.

• Convert to grayscale: The foundation for most document color adjustments

  import cv2
  import numpy as np
  
  img = cv2.imread("your_document.jpg")
  gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
  

• Simple binary thresholding: Works for evenly lit documents

  _, binary_img = cv2.threshold(gray, 127, 255, cv2.THRESH_BINARY)
  

• Adaptive thresholding: Way better for documents with uneven lighting/shadows

  adaptive_binary = cv2.adaptiveThreshold(
      gray, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, 11, 2
  )
  

  The Gaussian adaptive method calculates thresholds for small tile regions, so shadows don't wash out text.

Want to mimic the look of old, yellowed papers? This filter adds a warm, brownish tint:

# Start with grayscale, convert back to BGR to apply color tweaks
sepia = cv2.cvtColor(gray, cv2.COLOR_GRAY2BGR)
# Apply sepia color matrix
sepia[:, :, 0] = sepia[:, :, 0] * 0.272 + sepia[:, :, 1] * 0.534 + sepia[:, :, 2] * 0.131  # Blue channel
sepia[:, :, 1] = sepia[:, :, 0] * 0.349 + sepia[:, :, 1] * 0.686 + sepia[:, :, 2] * 0.168  # Green channel
sepia[:, :, 2] = sepia[:, :, 0] * 0.393 + sepia[:, :, 1] * 0.769 + sepia[:, :, 2] * 0.189  # Red channel
# Make sure values stay within 0-255 (uint8 range)
sepia = np.clip(sepia, 0, 255).astype(np.uint8)

If your scan is faded or has low contrast, use these methods to make text pop:

• Global histogram equalization: Quick but can over-brighten light areas

  equalized = cv2.equalizeHist(gray)
  

• CLAHE (Contrast Limited Adaptive Histogram Equalization): Recommended for documents
  It enhances contrast locally without blowing out bright regions:

  clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8,8))
  enhanced_doc = clahe.apply(gray)
  

  Adjust clipLimit (higher = more contrast) and tileGridSize based on your document's size.

Flip colors to turn white-background text into black-background—great for low-light reading:

# Invert full color image
inverted_color = cv2.bitwise_not(img)
# Invert grayscale image
inverted_gray = cv2.bitwise_not(gray)

Tweak the BGR channels (OpenCV uses BGR by default) to add warm or cool tones:

• Warm tone: Boost red, reduce blue

  warm_img = img.copy()
  warm_img[:, :, 0] = warm_img[:, :, 0] * 0.8  # Dim blue channel
  warm_img[:, :, 2] = warm_img[:, :, 2] * 1.2  # Brighten red channel
  warm_img = np.clip(warm_img, 0, 255).astype(np.uint8)
  

• Cool tone: Boost blue, reduce red

  cool_img = img.copy()
  cool_img[:, :, 0] = cool_img[:, :, 0] * 1.2  # Brighten blue channel
  cool_img[:, :, 2] = cool_img[:, :, 2] * 0.8  # Dim red channel
  cool_img = np.clip(cool_img, 0, 255).astype(np.uint8)
  

Quick Tip #

Remember OpenCV reads images in BGR format, not RGB. If you're using matplotlib to display results, convert it first with cv2.cvtColor(img, cv2.COLOR_BGR2RGB)—otherwise colors will look off!

内容的提问来源于stack exchange，提问作者EAS