Speed-up image processing on mobile CPU and GPU

As GPU becomes an integrated component in handheld devices, this paper introduces a speed-up image processing method utilizing low-power GPU and CPU in mobile platforms. When using GPU for general-purpose computing as well as CPU, data format conversion and data transfer to the other processor unit should be conducted. Moreover, idle time may occur whenever a processor unit is deliberately kept idle in the face of waiting tasks of the other processor unit. Considering these limitations, video decoding and data format conversion tasks are distributed to CPU side and image processing task is assigned to GPU side. These tasks of CPU and GPU are performed in parallel. Furthermore, to balance different capabilities of the processors, buffer control is implemented between the processors. The experimental results show that the proposed method can achieve significant speed-up compared to CPU and GPU sequential processing structure.