I. Introduction
The demand for portable multimedia devices that support video recording or video processing, such as smart phones and video cameras, is increasing. In order to reduce the memory consumed for video storage, the video recording system must contain video compression techniques. However, use of these techniques results in high computational complexity and external memory bandwidth requirement, causing significant power consumption. Video processing also involves a large number of computations and an increase in external memory traffic due to the usage of high-complexity vision algorithms such as deep neural networks. While many of these applications operate on battery power [1], [2], power management for video recording and video processing becomes a critical issue. Especially, as the resolution of video frames increases, the size of raw video data and the corresponding memory bandwidth increases as well. Thus, a frame memory compression (FMC) method based on embedded compression (EC) is becoming increasingly important for reducing the external data bandwidth requirement in video recording and video processing systems [3]–[8]. Note that the amount of the power consumed as a result of data transfer to external memory is very significant [9], although the internal power consumption can be greatly reduced by optimizing the operating conditions [10].