Old Photo Restoration

Overview

This skill enables AI-powered restoration and enhancement of old, damaged, or faded photographs using advanced image processing techniques and AI models.

When to Use This Skill

Use this skill when the user requests:

• Restoration of old or vintage photographs
• Enhancement of faded or discolored images
• Face recognition and enhancement in old photos
• Color correction or colorization of black and white photos
• Repair of scratches, tears, or other damage in photographs
• General improvement of image quality in vintage photos

Workflow

Step 1: Analyze the Photo

First, assess the photo's condition:

1. Identify issues:
• Fading or discoloration
• Scratches, tears, or physical damage
• Blurriness or noise
• Missing parts or torn edges
• Color casts (yellowish, reddish tones)

1. Detect faces:
• Count the number of faces
• Identify face positions and orientations
• Assess face quality (occlusion, blur, lighting)
• Note any facial damage or distortion

1. Determine restoration needs:
• Is colorization needed? (black and white photos)
• Is facial enhancement required?
• What level of detail restoration is needed?
• Are there specific areas requiring special attention?

Step 2: Choose Enhancement Approach

Based on analysis, select appropriate techniques:

For Color Issues:

• Use color correction algorithms
• Apply histogram equalization
• Adjust white balance
• For black and white photos, consider AI-based colorization

For Facial Enhancement:

• Apply face detection using OpenCV or similar
• Use AI-based face restoration models (e.g., CodeFormer, GFPGAN)
• Enhance facial features while preserving identity
• Smooth skin texture naturally

For Damage Repair:

• Use inpainting algorithms for small scratches
• Apply denoising filters
• Use content-aware fill for missing areas
• Blend restored areas seamlessly

Step 3: Execute Restoration

1. Preprocessing:
• Normalize image brightness and contrast
• Apply basic denoising
• Correct geometric distortions if present

1. Face Enhancement (if faces detected):
• Detect all faces in the image
• Extract face regions
• Apply face restoration model
• Blend enhanced faces back into original image

1. Color Correction:
• Analyze color distribution
• Remove unwanted color casts
• Enhance contrast and saturation
• Preserve natural skin tones

1. Detail Enhancement:
• Apply sharpening filters appropriately
• Enhance texture details
• Reduce noise without losing important details

Step 4: Post-Processing

1. Review the result:
• Compare before and after
• Check for artifacts or unnatural enhancements
• Ensure faces still look natural and recognizable

1. Fine-tune adjustments:
• Adjust enhancement intensity if needed
• Balance between restoration and preservation of vintage feel
• Apply final touches (cropping, slight brightness adjustments)

1. Output formats:
• Save in high-quality format (PNG or high-JPEG quality)
• Consider saving both before and after versions
• Document the restoration process if requested

Using the Script

This skill includes a Python script for automated photo restoration:

scripts/restore_photo.py <input_image> [options]

Options:

• --enhance-faces: Enable facial enhancement using AI models
• --color-correct: Apply automatic color correction
• --colorize: Convert black and white to color (experimental)
• --remove-scratches: Attempt to remove small scratches and damage
• --output : Specify output file path
• --intensity <0-1>: Enhancement intensity (default: 0.7)

Examples:

Basic restoration with face enhancement:

python scripts/restore_photo.py old_photo.jpg --enhance-faces --output restored_photo.jpg

Full restoration with all features:

python scripts/restore_photo.py vintage.png --enhance-faces --color-correct --remove-scratches --intensity 0.8

Colorize black and white photo:

python scripts/restore_photo.py bw_photo.jpg --colorize --output colorized.jpg

Resources

scripts/

restore_photo.py - Main restoration script that:

• Loads and analyzes input images
• Detects faces using OpenCV
• Applies AI-based face restoration
• Performs color correction and enhancement
• Outputs restored images with customizable options

Requirements: The script requires Python packages:

• OpenCV (cv2)
• NumPy
• Pillow (PIL)
• Optional: face restoration model dependencies

references/

image_processing_techniques.md - Detailed documentation on:

• Image preprocessing methods
• Color correction algorithms
• Face detection and enhancement techniques
• Best practices for photo restoration

model_installation.md - Guide for installing AI models:

• Face restoration model setup
• Model download instructions
• Performance optimization tips

assets/

This directory is not needed for this skill as all processing is handled programmatically.

Best Practices

1. Preserve Originality: Always keep the original image file. Restoration should enhance, not erase, the photo's history.

1. Natural Enhancement: Avoid over-enhancement. Subtle improvements often produce better results than aggressive processing.

1. Test Incrementally: Apply changes step-by-step and review after each step to ensure the desired effect.

1. Face Recognition Priority: For photos with identifiable people, prioritize maintaining facial features and expressions over other enhancements.

1. Quality Trade-offs: Balance between enhancement quality and processing time. Higher quality may require more computational resources.

1. Multiple Versions: Consider creating different versions with varying enhancement levels, allowing the user to choose the best result.

Common Scenarios

Scenario 1: Faded Family Photo

• Detect faces for enhancement
• Apply color correction to remove yellowing
• Increase contrast subtly
• Preserve original composition

Scenario 2: Black and White Portrait

• Use AI colorization
• Enhance facial details
• Ensure natural skin tones
• Maintain vintage character

Scenario 3: Damaged Photo with Scratches

• Remove small scratches via inpainting
• Enhance damaged areas
• Apply overall denoising
• Blend restored sections carefully

Scenario 4: Multiple Group Photo

• Detect and enhance all faces
• Balance lighting across faces
• Apply uniform color correction
• Ensure consistent enhancement

Limitations

• Face recognition works best with frontal faces. Profile or partially occluded faces may have limited enhancement.
• Extreme damage (large tears, missing sections) may require manual editing beyond automated tools.
• Colorization accuracy depends on AI model training and may not always match historical reality.
• Very low-resolution images may have limited enhancement potential.
• Over-enhancement can create unnatural artifacts; use conservative intensity settings.

Troubleshooting

No faces detected: Check image quality and lighting. Ensure faces are visible and not extremely small in the frame.

Unnatural colors: Reduce color correction intensity or manually adjust white balance parameters.

Over-smoothed faces: Lower the enhancement intensity to preserve natural texture and details.

Artifacts in restored areas: Reduce overall enhancement intensity or adjust specific parameters for problematic areas.

Processing too slow: For large images, consider resizing before processing or reduce enhancement intensity.

Notes

• Always obtain consent before enhancing photos of people, especially for sharing or public use.
• Consider the historical context of vintage photos; sometimes preserving age signs is more appropriate than removing them completely.
• Batch processing of multiple photos is supported through the script's command-line interface.
• The restoration quality depends on the original image quality and the extent of damage.