CRA: A Real-Time CUDA-Based Resolution Adapter in Screen Recording Applications

Recently, screen recording technology is widely used by lots of applications, such as remote control, distance learning, and distance conference etc. A problem, called Asymmetric Screen Resolution (ASR) in this paper, will occur in these applications. The ASR problem means that the resolution of a recording computer is different from that of a display computer. Different computer system will have different screen resolution and the ASR problem will result in screen recording resolution mismatch between the recording computer and display computer. This paper proposes a novel CRA algorithm to apply advantage of the CUDA parallel computing for adjusting the resolution on the Graphics Processing Unit (GPU). This algorithm effectively reduces CPU utilization and guarantees the display quality after adjusting resolution. Four different application scenarios, web browsing, text editing, slide presentation, and video playback are used to test the performance of our proposed CRA algorithm. These experiments measure CPU utilization and the PSNR of image adjusted by the CRA algorithm. The experimental results show that CRA algorithm effectively reduce CPU utilization 32%~40% while using whole screen in resolution adjustment. The CRA algorithm effectively reduces CPU utilization 0.1%~13% while using difference rectangle in resolution adjustment. On image quality measurements, the PSNR value is controlled between 25db~28db while using the CRA algorithm to adjust screen resolution from 1920×1080 to 1536×864. The PSNR value is controlled between 26db~34db by using the CRA algorithm to adjust screen resolution from 1920×1080 to 960×540. Using CRA algorithm presented in this paper will definitely get a good image quality, and effectively reduce CPU utilization.