Performance and power consumption evaluation in smartphone based image processing for medical applications

More and more image processing applications are intended to run on mobile devices. This trend is visible also in medical industry, which have very strict reliability and safety requirements. Increased power consumption and inefficient image processing can affect both of these factors significantly. More over, it affects the user experience as well. In this paper authors investigates the performance and power consumption of several different numeric algorithms implemented on iPhone device. The comparison is made between CPU and GPU based solutions. Authors evaluated the test scenarios involving common used operations: matrix multiplication, thresholding and Canny's edge detection. The results showed that GPU based solution can be up to 4 times faster as CPU based one, with similar power footprint in particular cases. All measurements have been gathered without hardware interference nor jailbreaking of the device. Finally authors discuss possible application considerations and scenarios in medical related image processing smartphone apps.