antoinevalera/h3d-cf

3D Color Homography Model for Photo-Realistic Color Transfer Re-Coding, TVC 2017

h3d-cf

The code is provided for the following publications:

[1] Gong, H., Finlayson, G.D., Fisher, R.B. and Fang, F., 2017. 3D color homography model for photo-realistic color transfer re-coding. The Visual Computer, pp.1-11.

[2] Gong, H., Finlayson, G.D., Fisher, R.B.: Recoding color transfer as a color homography. In: British Machine Vision Conference. BMVA (2016)

Please cite the above papers if you find them useful for your research.

Copyright 2018 Han Gong ([email protected]), University of {East Anglia, Edinburgh}.

Introduction

Color transfer is an image editing process that naturally transfers the color theme of a source image to a target image. In this paper, we propose a 3D color homography model which approximates photo-realistic color transfer algorithm as a combination of a 3D perspective transform and a mean intensity mapping. A key advantage of our approach is that the re-coded color transfer algorithm is simple and accurate. Our evaluation demonstrates that our 3D color homography model delivers leading color transfer re-coding performance. In addition, we also show that our 3D color homography model can be applied to color transfer artifact fixing, complex color transfer acceleration, and color-robust image stitching.

Usage

• cf_3D_H.m: the code of the TVC paper [1].
• cf_2D_H.m: the code of the BMVC paper [2].

Testing

• test_main.m: this file contains the evaluation code for re-producing the results in the paper.

cf_MK.m and cf_Poly.m are the implementations for two other color transfer approximation methods for comparison. The approximation outputs will be saved in the folder out_pair.

Dataset

A set of 200 transfer cases for 4 different color transfer methods (800 test cases in total) are contained in the folder in_pair. Note that the code for the original color transfer methods is not provided here. Only the cached results for color transfer approximation are provided in the dataset.

Rules of Filenames:

• *_s.jpg: original source image.
• *_s.jpg: original target image.
• *_[method_name].jpg: color transfer result produced by [method_name].

Applications

The proposed method may be used for (see the original paper for higher-quality images):

• color transfer artifact fixing
• complex color transfer acceleration
• color-robust image stitching.