Image Enhancer

This is a self-contained skill. Do NOT load external files or reference directories.

Mindset & Philosophy

A professional screenshot or image in documentation represents the quality of the product itself. The goal of enhancement is maximizing clarity, readability of text, and visual appeal while minimizing file size. Do not rely on generic upscalers. Apply precise, surgical digital image processing techniques based on the source characteristics.

---

Tool Selection & Decision Tree

Evaluate the image properties (resolution, type, format) and apply the correct approach:

graph TD
    A[Image Input] --> B{What is the content type?}
    B -->|UI Screenshot with Text| C{Has blur or low res?}
    B -->|Photo or Complex Illustration| D{Has noise or artifacts?}
    
    C -->|Yes| E[Upscale 2x using Nearest Neighbor & Sharpen via ImageMagick]
    C -->|No| F[Lossless compression via optipng]
    
    D -->|Yes| G[Denoise using Python OpenCV bilateralFilter]
    D -->|No| H[Lossy optimization via jpegoptim]

Choice Matrix

Scenario	Primary Tool	Fallback / Alternative	Key Parameter Strategy
UI/Terminal Screenshot	convert (ImageMagick)	Python (Pillow)	Use nearest-neighbor scaling (-scale 200%) to keep text pixel-perfect. Avoid bilinear scaling.
Blurry UI Text	convert	Python (Pillow)	Apply sharpening filter (-unsharp 0x1+1.0+0.05 or -sharpen 0x1.5).
Photo/Denoising	Python (cv2 bilateralFilter)	Python (scipy median filter)	Preserves edges while smoothing flat areas.
PNG Compression	optipng	pngquant	Lossless (optipng -o7) vs lossy PNG quantizing (pngquant --quality=80-90).
JPG Compression	jpegoptim	convert	Strip metadata (--strip-all) and target specific quality (-m85).

---

Domain-Specific Procedures & Commands

1. Enhancing UI Screenshots (Pixel-Perfect Text)

Standard resizing introduces blur. Use nearest-neighbor interpolation to double size, then apply a subtle unsharp mask to crisp the edges:

# Keep text crisp while upscaling UI screenshots
convert input.png -scale 200% -unsharp 0x0.75+0.75+0.008 output.png

2. Lossless Screenshot Size Reduction (Critical for Page Load)

Never ship raw screenshots. Compress losslessly to preserve alpha and visual quality:

# Level 7 lossless optimization (removes unnecessary chunks & metadata)
optipng -o7 -strip all input.png -out optimized.png

3. Edge-Preserving Denoising for Blurry/Compressed Images

If using a custom python script, prioritize Bilateral Filtering to clean background noise without blurring key edges:

import cv2

# Load image
img = cv2.imread('input.jpg')
# d=9 (neighborhood diameter), sigmaColor=75, sigmaSpace=75
denoised = cv2.bilateralFilter(img, 9, 75, 75)
cv2.imwrite('output.jpg', denoised)

---

NEVER Anti-Patterns

Action	Why	Consequences	Correct Alternative
NEVER use default -resize in ImageMagick for pixel art or UI screenshots.	Standard lanczos/cubic resize interpolates pixels and introduces blur around text.	Blurry text, hard-to-read documentation.	Use -scale (nearest-neighbor) for exact percentage increments.
NEVER save UI screenshots as JPG.	JPG compression creates ringing artifacts around high-contrast text edges.	Ugly color noise, unreadable text, bloated file size.	Always use PNG format for screenshots.
NEVER overwrite original images.	Destructive editing cannot be undone if the parameter choice was sub-optimal.	Loss of source files, manual rework.	Always output to a new file suffix (e.g. _enhanced.png) and retain the source.
NEVER leave EXIF/GPS metadata in web images.	Exposes sensitive user information (location, device specs, creation dates).	Privacy leak, security violation.	Use -strip in ImageMagick or --strip-all in jpegoptim.

---

Freedom Calibration

• Low Freedom (Strict Compliance): For final output format optimization (must strip metadata, must use PNG for text screenshots, must preserve original as backup).
• Medium Freedom (Operational Guidance): Sharpening parameters (-unsharp values) and upscaling percentages can be tweaked based on visual checks and zoom level.

---

Error Mitigation & Failure Modes

1. Tool Missing Errors

If CLI tools are not installed, fallback gracefully:

• No optipng: Fallback to python Pillow library to save with high optimization levels:

  from PIL import Image
  img = Image.open('input.png')
  img.save('output.png', optimize=True)
  

• No ImageMagick convert: Check if gm (GraphicsMagick) is installed, otherwise run a Python script using Pillow with Resampling.NEAREST or Resampling.LANCZOS filters.

2. Output File Exceeds Original Size

Sometimes aggressive optimization or upscaling makes a file larger than the original.

• Action: Always run a size check. If the "enhanced" file is larger, and no significant visual gains were achieved, revert to the original or apply quantization (pngquant or lower quality setting).

6) Capture Knowledge

After an image is enhanced, optimized, or converted for documentation, automatically trigger the capture_knowledge.py script. The script will analyze the image processing steps to identify:

• New image enhancement patterns (e.g., specific -unsharp values for UI text).
• Optimization results (size reduction ratios, format changes).
• Specific tools and versions used for the enhancement. The script will then route this information to the appropriate storage:
• OKF: Image processing standards (e.g., "always use PNG for screenshots," "default upscaling values").
• ChromaDB: Specific enhancement parameters (e.g., "UI screenshot sharpen: -unsharp 0x0.75+0.75+0.008").