Fast color transfer for camouflage applications

Abstract

Color transfer techniques have been widely developed to manipulate color information of image data in many applications. In this paper, color transfer research is presented for camouflage applications, along with the challenges facing many state-of-the-art color transfer methods. To address these challenges, a new fast color transfer algorithm is proposed. This new fast color transfer algorithm follows a classical color transfer framework, consisting of two parts: domain color extraction and color replacement. However, unlike other existing color transfer algorithms, this algorithm performs domain color extraction using a sampling process to reduce the dataset size of the classification process within the algorithm. Once the dataset size of the information to be classified is reduced, corresponding computational cost is largely decreased, thus improving the speed of the color transfer algorithm. Following previous domain color extraction process to collected domain color information, color replacement is conducted to change the color distribution of the target images (camouflage images) to be similar to those of the source images (operational background images). The entire algorithm is processed in a luminance-chrominance-independent (Lαβ) color model space. Experimental results show the performance of the proposed algorithm.