The identification of individuals’ natural hair color is very important for proper treatment and coloring. Hence, ByBente has a color system which is applied to describe the color composition of a given individual’s hair. According to the system, a color of an individual’s hair can be labeled with a number like 7.23, in which individual numbers corresponds to a particular color attribute of the person’s hair. Different hair color product recommendation has been given at ByBente depending on such identification of individual’s hair color composition.
Currently, the identification of hair color composition is being performed by a human expert. ByBente has been working on automating the process and design a recommender system for hair color customers. Like prior and related color imaging applications, hair color identification can be automated with efficient computer vision and machine learning approaches. For more effective, remote, as well as convenient applicability of the system, smartphone-based solution is considered. However (related to color imaging applications) the material composition of hair, the illumination condition, as well as the uncontrolled acquisition process of hair images will influence the appearance of the hair color. This variability of hair color is expected to create a challenge on the automation of the identification system. Therefore, various color management and image enhancement approaches will need to be included.
Accordingly, this pre-project is set to investigate efficient and state of the art classification and identification techniques for the intended hair color recommender system. Color fidelity issues, which arises with smartphone-based applications, will also be thoroughly analyzed. Finally, a recommendation for hair color classification and identification algorithms, pre-processing and enhancement techniques, as well as representative data set structure will be offered.