基於視覺顯著度的低運算成本省電方法

Abstract

行動裝置的使用在近年來逐漸增加，而螢幕是行動裝置中耗電占比大的部分之一，在使用行種裝置時，需開啟螢幕來執行各項操作像是觀看影片和撰寫訊息等，這些情境使得螢幕的耗電情況不容忽視。為了降低螢幕的耗電量並兼顧使用者體驗，有研究針對應用程式介面提出省電架構，配合視覺顯著圖並利用元件檔案將圖片分割為多個區域，將各個區域以更低功耗的色彩呈現以達省電效果，然而此架構卻有數百秒的運算時間。本論文提供6個方法更改上述省電架構，使省電架構在同樣的條件限制下有更短的運算時間。實驗結果顯示，本研究使省電架構的運算時間範圍從278秒至731秒，下降至介於45秒至137秒。

The use of mobile devices has gradually increased in recent years, and the screen is one of the major power-consuming components in these devices. When using mobile devices, the screen needs to be turned on for various operations such as watching videos and composing messages. These scenarios make the power consumption of the screen significant. In order to reduce the power consumption of the screen and maintain user experience. There have been studies proposing a power-saving architecture for application programming interfaces, which involves using visual saliency maps and component files to divide the screen into multiple regions. Each region is presented with colors that consume lower power. However, this architecture has a computation time of several hundred seconds. This paper presents six methods to modify the aforementioned power-saving architecture, aiming to reduce the computation time under the same constraints. The experimental results show that this research has reduced the computation time of the power-saving architecture from a range of 278 to 731 seconds to a range of 45 to 137 seconds.