Color Descriptor for Mobile Augmented Reality

Abstract

Augmented Reality (AR) is the combination of real-world and virtual content where it allows users to interact with virtual objects in real-time. In order to achieve optimal performance in mobile AR applications, choosing a suitable tracking algorithm is an important issue as the smartphone has its own limited processing capabilities and memories when compared to the PC platform. Tracking in AR consists of four important components; detector, descriptor, matcher and pose estimator. The robustness of mobile AR application in illumination is an issue that has not been resolved. The descriptor is one of the important components as it will affect the overall performance of mobile AR applications. In order to create an efficient and robust mobile AR application, the descriptor must be efficient which means that it must be able to function in a short time and robust to changes in illumination. FREAK (Fast Retina Keypoint) has been proposed as the most appropriate descriptor for mobile AR applications based on previous research. FREAK can function in the shortest time and most robust to changes in scale and rotation. Although FREAK can function in the shortest time and robust to changes but the color information that will affect the illumination invariance of a descriptor had been ignored by FREAK. This study has developed a FREAK descriptor using RGB color space to improve the robustness of illumination invariance. A vertical concatenation technique has been used to combine all the key points extracted. Evaluation for the robustness of descriptors in terms of illumination changes has been done using the ALOI data set. The accuracy of the RGB FREAK descriptor has increased in illumination invariance (light color changes and lighting arrangement changes) compared to the original FREAK descriptor.