Color selection plays an important role in drawing based image retrieval. Frequently, users are overloaded by color options to select while drawing the query. In this paper, a color recommendation method dedicated to drawing-to-retrieve image search is proposed and implemented in our draw-to-retrieve application, which aims at assisting users in choosing colors in order to improve image retrieval efficiency in terms of drawing time and retrieval accuracy. Then, this method is tested in a preliminary experiment designed to explore how the image retrieval efficiency is affected when different colors are provided the palette. In order to illustrate an application of proposed approach, images of natural scenes are used as an example. 23 participants took part in the experiment to perform draw-to-target tasks of 36 images using three different color palettes. There are mainly three results concluded from the experiment: firstly, the image retrieval efficiency drops as the difficulty of target image increases; in addition, providing useful colors in the palette can lead to an more efficient retrieval; finally, the retrieval has better performance using our color recommendation approach than commonly used 20-color palette and a subjective evaluation of the participants is also included.
Zhan Xu, Chao Zhang, Guoping Qiu, "Color Recommendation for Drawing Based Image Retrieval on Mobile Devices" in Proc. IS&T 24th Color and Imaging Conf. , 2016, https://doi.org/10.2352/ISSN.2169-2629.2017.32.254