Mist Effect Pipeline

This project creates artistic mist/fog and Orton effects on photographs using Depth Anything V2 for depth estimation and depth-dependent processing.

How it Works

1. Input: Takes a photograph (JPG, TIFF, PNG, etc.)
2. Depth Estimation: Uses Depth Anything V2 to generate a depth map
3. Mist Effect: Applies depth-dependent contrast reduction - distant objects get more contrast reduction, creating a natural mist/fog effect
4. Variable Orton Effect: Applies depth-aware blur and overlay effects that respect depth ordering - distant objects get more blur/overlay, but close objects are never blurred by distant objects behind them
5. Output: Saves the processed image with the combined effects

Installation

1. Create a virtual environment:

python3 -m venv venv
source venv/bin/activate

2. Install dependencies:

pip install -r Depth-Anything-V2/requirements.txt

3. Download the Depth Anything V2 model checkpoint:

# The checkpoint file should be placed in the root directory
# depth_anything_v2_vitl.pth (already present in this repository)

Usage

Single Image Processing

python mist_effect.py --input path/to/image.jpg --output misty_image.jpg

Batch Processing

python mist_effect.py --input-dir path/to/images --output-dir path/to/output

Parameters

• --contrast-reduction: How much to reduce contrast for distant objects (0.0-1.0, default: 0.5)
• --brightness-adjustment: Brightness adjustment for distant objects (-1.0 to 1.0, default: 0.0)
• --orton-blur: Maximum blur strength for Orton effect (0.0-1.0, default: 0.0)
• --orton-overlay: Maximum overlay opacity for Orton effect (0.0-1.0, default: 0.0)
• --encoder: Depth Anything V2 encoder ('vits', 'vitb', 'vitl', 'vitg', default: 'vitl')
• --input-size: Input size for depth estimation (default: 518)

Examples

# Basic mist effect
python mist_effect.py --input photo.jpg --output misty_photo.jpg

# Strong mist effect
python mist_effect.py --input photo.jpg --output misty_photo.jpg --contrast-reduction 0.8 --brightness-adjustment 0.1

# Orton effect only
python mist_effect.py --input photo.jpg --output orton_photo.jpg --orton-blur 0.6 --orton-overlay 0.3

# Combined mist and Orton effects
python mist_effect.py --input photo.jpg --output combined_photo.jpg --contrast-reduction 0.3 --orton-blur 0.4 --orton-overlay 0.2

# Process all images in a directory
python mist_effect.py --input-dir ./photos --output-dir ./processed_photos

# Use different encoder for faster processing
python mist_effect.py --input photo.jpg --output misty_photo.jpg --encoder vits

Effect Parameters

• Contrast Reduction (0.0-1.0):

  • 0.0 = No contrast reduction (no mist effect)
  • 0.5 = Moderate mist effect (default)
  • 1.0 = Maximum contrast reduction (strong mist effect)

• Brightness Adjustment (-1.0 to 1.0):

  • 0.0 = No brightness change (default)
  • Positive values = Brighten distant objects
  • Negative values = Darken distant objects

• Orton Blur Strength (0.0-1.0):

  • 0.0 = No blur (default)
  • 0.5 = Moderate blur for distant objects
  • 1.0 = Maximum blur for distant objects

• Orton Overlay Opacity (0.0-1.0):

  • 0.0 = No overlay effect (default)
  • 0.3 = Subtle overlay with brightness and saturation boost
  • 1.0 = Maximum overlay effect
  • Note: Uses "lighten only" blending - only brightens, never darkens

Technical Details

Mist Effect

The mist effect is achieved by:

1. Depth Normalization: The depth map is normalized to 0-1 range
2. Contrast Reduction: For each pixel, contrast is reduced based on its depth value
3. Brightness Adjustment: Optional brightness adjustment for distant objects
4. Channel Processing: Each color channel (R, G, B) is processed independently

The algorithm uses the formula:

processed_pixel = mean + (original_pixel - mean) * contrast_factor

Where contrast_factor = 1.0 - (depth_normalized * contrast_reduction_factor)

Variable Orton Effect

The Orton effect is achieved by:

1. Sharp Base Image: Keeps the original image crisp as the foundation
2. Single Blur Pass: Creates one blurred version of the entire image
3. Depth-Aware Blending: Blends blurred and sharp versions based on depth
4. Natural Depth Ordering: Distant objects get more blur, close objects stay sharp
5. Lighten-Only Overlay: Brightness and saturation enhancement that only brightens, never darkens

Key Benefits:

• No bleeding: Dark areas never bleed into light areas
• Natural depth-of-field: Mimics real camera behavior
• Crisp foregrounds: Close objects remain sharp
• Ethereal backgrounds: Distant objects get soft, dreamy appearance

This creates a natural depth-aware effect where distant objects appear softer and more ethereal while maintaining sharpness in foreground elements.

Requirements

• Python 3.7+
• PyTorch
• OpenCV
• NumPy
• Matplotlib
• PIL/Pillow

Model

This project uses Depth Anything V2 (ViT-L encoder) for depth estimation. The model checkpoint should be placed in the root directory as depth_anything_v2_vitl.pth.

depth-mist