Comparative Study of Image Processing Performance of Camera-Based Visible Light Communication Using Android Acceleration Frameworks

The asynchronous nature of smartphone-to-smartphone (S2S) based on visible light communication (VLC) imposes a significant challenge on the speed of camera-based receiver processing time and algorithm. Recent improvements on the smartphone camera hardware and the current release of the highly customised camera2 application programming interface (Camera2-API) have increased the smartphone’s computational capability. This paper presents a comparative study of the acceleration frameworks, which can be used for image processing on Android device to maximize the code performance, thus reducing the computational time of data frame detection. An experimental S2S VLC system is developed for evaluation of the graphical processing unit acceleration (GPU), Android runtime (ART) and native development kit (NDK) based algorithms for processing the captured data. In addition, we determine the total number of processed pixels for multiple frames with the maximum possible detection frequency for S2S VLC. Using the additive property of RGB colour space, two sets of experiments are implemented: firstly the conversion from YUV to RGBA (Red Green Blue Alpha) using all of the available colour-based data, which leads to ~500% of improvement in colour conversion time using NDK compared to ART. A gain of 200% is also achieved compared to GPU-based algorithms. Secondly, the grayscale filtered YUV to RGBA conversion shows that NDK processing time is 200% faster than the direct ART, which outperforms GPU conversion at lower frame sizes. From the results findings, we propose an optimal approach for camera-based VLC application development using Android smartphones.