I. Introduction
Power saving is a very important technique now, because of the rapid development of mobile and portable devices whose battery only has limited stored power. Therefore the power saving mechanism is crucial in the electronic system. Take smart phone for example, the TFT-LCD on it can consume up to 20-45% of total power usage depending on the different operating mode. Thus the backlight power saving can contribute to the large portion of the total saving. Backlight is a technique used for liquid crystal display (LCD) to show image/video pixels, by using light sources from the back or sides of the LCD, to modulate the brightness. Higher brightness introduces higher energy consumption. Thus, an intuitive way to lower the energy consumption is to lower the brightness range of the pixels. But this distorts the quality of the images and videos. There are existing methods in backlight power saving to control the luminance and contrast to reduce the power consumption. To reduce the decrease of the image quality due to power saving, the work in [1] proposed a new backlight dimming algorithm and a new image enhancement algorithm to consider the power saving and the image quality at the same time. In [2], the method finds the maximum pixel value of the block, and reduces the value by 50% to save energy. In [3], authors consider the distributions of R, G and B layers to perform backlight saving algorithm. In [4], the maximum luminance of the processed image is considered for image integrity at a desired level. The work in [5] proposed multi-histogram-based backlight dimming algorithm utilizing different pixel distributions in an image. In [6], K-means clustering method is used to find segmentations in an image. The pixel distributions of different segmentations are considered to find best clipping point. The computation complexity is reduced in [7]. Backlight dimming algorithm focusing on viewing distances and color distortion are discussed in [8] and [9], respectively. For more recent technology in display, Organic Light-Emitting Diode (OLED) display becomes more popular due to its efficiency. In [10], the method uses an objective function to consider the tradeoff between image quality and power consumption, and a sampling method is used to reduce the complexity. The work in [11] makes uses of the SSIM (The Structural SIMilarity index [12]) to design a perceptual quality aware power reduction method for OLED display.