A new class of very efficient algorithms for local dimming

One of the main aspects of today’s computing, especially on mobile devices, is power consumption. It affects the lifetime of batteries and has ecological aspects. In the near future, a significant proportion of the energy of mobile devices will be spent on displays. Thus, dimming, especially local dimming of displays, increases the comfort of these mobile devices. A convenient side-effect of local dimming is contrast enhancement and a better black level. Local dimming has three main aspects: the image processing aspect, the optimization aspect of the core algorithm and real-time requirements. We deal with the optimizer part, also focusing on real-time aspects. In this article, a new class of generalized dimming algorithms GDA(n) is developed. This class depends on a single parameter, allowing us to steer between power efficiency and smoothness of the solution. The smoothness properties of the proposed algorithms allow them to be treated as regularizations of the sorted sector covering (SSC) algorithm. The SSC algorithm forms a foundation for our algorithms, and it will be described later in this article. The implementation of the proposed algorithms is quite simple, e.g. on the basis of an existing implementation of the SSC, and they are highly effective. Most important, their smoothness is an inherent part of the algorithm, thus reducing flickering effects before they are created. Numerical examples comparing GDA(n) to established algorithms are given, substantiating the efficiency and quality of the new method. By steering the parameter n, we can switch from a smooth distribution of the LED values (n small) to a volatile distribution of the LED values (n large), while preserving the required brightness of the backlight of the display. In the first case, we suppress LED flashlighting and, especially for videos, flickering. In the latter case we adapt the LED backlight better to the image brightness, obtaining dark LED values in dark areas of the image and bright LED values in bright areas of the image.