colortrans

An implementation of various algorithms for transferring the colors from a reference image to a content image while preserving the qualitative appearance of the content image (i.e., color transfer).

Installation

Requirements

• Python 3.8 or greater (earlier versions may work, but are not tested)

Install

$ pip3 install colortrans

Update

$ pip3 install --upgrade colortrans

Command-Line Usage

The program can be used from the command line.

The general command line usage is shown below.

$ colortrans [--method METHOD] [--single-precision] CONTENT REFERENCE OUTPUT

CONTENT is the path to the content image, REFERENCE is the path to the style image, and OUTPUT is the path to save the output image.

METHOD specifies the color transfer algorithm. The following methods are supported:

1. lhm Linear Histogram Matching [1] (default)
2. pccm Principal Components Color Matching [2, 3]
3. reinhard Reinhard et al. [4]

If the optional --single-precision flag is present, 32-bit floats will be used instead of 64-bit floats.

If the launcher script was not installed within a directory on your PATH, colortrans can be launched by passing its module name to Python.

$ python3 -m colortrans [--method METHOD] CONTENT REFERENCE OUTPUT

Library Usage

The algorithms can also be used directly from Python programs. Each of the methods listed above has a corresponding function, transfer_METHOD, taking two NumPy arrays corresponding to the content and reference image, respectively. The arrays have HxWxC data ordering (channels-last). The functions also take an optional keyword argument, single_precision, a boolean that specifies whether 32-bit floats will be used instead of 64-bit floats (defaults to False).

Example

import colortrans
import numpy as np
from PIL import Image

# Load data
with Image.open('/path/to/content.jpg') as img:
    content = np.array(img.convert('RGB'))
with Image.open('/path/to/reference.jpg') as img:
    reference = np.array(img.convert('RGB'))

# Transfer colors using different algorithms
output_lhm = colortrans.transfer_lhm(content, reference)
output_pccm = colortrans.transfer_pccm(content, reference)
output_reinhard = colortrans.transfer_reinhard(content, reference)

# Save outputs
Image.fromarray(output_lhm).save('/path/to/output_lhm.jpg')
Image.fromarray(output_pccm).save('/path/to/output_pccm.jpg')
Image.fromarray(output_reinhard).save('/path/to/output_reinhard.jpg')

References

[1] Hertzmann, Aaron. "Algorithms for Rendering in Artistic Styles." Ph.D., New York University, 2001.

[2] Kotera, Hiroaki, Hung-Shing Chen, and Tetsuro Morimoto. "Object-to-Object Color Mapping by Image Segmentation." In Color Imaging: Device-Independent Color, Color Hardcopy, and Graphic Arts IV, 3648:148–57. SPIE, 1998.

[3] Kotera, Hiroaki. "A Scene-Referred Color Transfer for Pleasant Imaging on Display." In IEEE International Conference on Image Processing 2005, 2:II–5, 2005.

[4] Reinhard, Erik, Michael Adhikhmin, Bruce Gooch, and Peter Shirley. "Color Transfer between Images." IEEE Computer Graphics and Applications 21, no. 5 (July 2001): 34–41.