andersource / image-color-replacement

Image color replacement using numerical optimization

image-color-replacement

Image color replacement using numerical optimization.

Usage

• pip install -r requirements.txt
• Edit config.yml (configuration options explanation below)
• python image_color_replacement.py

Configuration

• input_image_path: Path of source (input) image
• output_image_path: Path to save target image
• intermeidate_representation_path (optional): Path to save intermediate image representations
• from_colors: List of RGB colors to replace
• to_colors: List of RGB colors with which to replace colors from from_colors, in matching order
• fixed_colors: List of RGB colors that shouldn't change
• latent_color_components_k: Number of latent color components
• optimization_scale: How much to downscale image before performing optimization (it is not a good idea to optimize on an unscaled large image)
• wolfe_conditions: Wolfe line search parameters (only used if BFGS implementation is "scratch")
• penalty_weights: List of penalty weights (weighting objective vs. constraint violation)
• bfgs_max_iterations: Max BFGS iterations (per penalty weight)
• bfgs_implementation: One of "scipy" and "scratch", "scipy" highly advised

More

See this post for more details.