Visualizing Large Datasets on Memory and Performance Constrained Mobile Devices

Graduation date: 2016Graphics hardware in mobile devices has become more powerful, allowing rendering techniques such as ray-cast volume rendering to be done at interactive rates. This increase of performance provides desktop capabilities combined with the portability of a tablet. Volumes can demand a high amount of memory in order to be loaded in. This becomes problematic when dealing with mobile operating systems, such as Android, while trying to load large volumes into an application. Even though tablets on the market today contain 1 – 3 gigabytes of memory, Android allocates only a fraction of the total memory per application. Cases in which the dataset does fit into memory, but the resolution of the volume surpasses the capabilities of the mobile GPU, results in an unresponsive application. Although downscaling the data is a remedy to both the lack of memory and GPU performance, it is sacrificing potentially useful information. This loss of data is undesired in scientific fields, such as medical imaging. Combining both downsizing and data division tactics, this research project introduces a method that allows the user to view the entirety of the dataset as a whole and zoom in on the native resolution sub-volumes. Additionally, our method allows the GPU to perform at an effective level to achieve interactive frame rates