Interactive time-dependent tone mapping using programmable graphics hardware

Abstract

Modern graphics architectures have replaced stages of the graphics pipeline with fully programmable modules. Therefore, it is now possible to perform fairly general computation on each vertex or fragment in a scene. In addition, the nature of the graphics pipeline makes substantial computational power available if the programs have a suitable structure. In this paper, we show that it is possible to cleanly map a state-of-the-art tone mapping algorithm to the pixel processor. This allows an interactive application to achieve higher levels of realism by rendering with physically based, unclamped lighting values and high dynamic range texture maps. We also show that the tone mapping operator can easily be extended to include a time-dependent model, which is crucial for interactive behavior. Finally, we describe the ways in which the graphics hardware limits our ability to compress dynamic range efficiently, and discuss modifications to the algorithm that could alleviate these problems.

Fast Local Tone Mapping, Summed-Area Tables and Mesopic Vision Simulation

“…Several attempts were made to accelerate this filtering to provide real-time frame rates. Goodnight et al (2003) investigated the implementation of local operators on GPU. Their best result was with the photographic operator, where they implemented the 2D Gaussian convolution using separable 1D kernels in a two-stage approach.…”

“…The process of SAT generation can be efficiently implemented on the GPU, thus not only mitigating the occurrence of halos, but also substantially accelerating the operator, even when compared against the fast method of Krawczyk et al (2005). A review on SAT and their generation is provided in Sections 2.4,3.2 and 3.3. (a) Reinhard et al (2002) (b) Goodnight et al (2003) (c) Krawczyk et al (2005) (d) Slomp & Oliveira (2008) This Chapter describes a more attractive method for SAT generation based on balanced-trees [Blelloch (1990)]. The originally employed recursive-doubling algorithm [Dubois & Rodrigue (1977)] requires less passes, but performs more arithmetic operations and is more sensitive to cache thrashing.…”

“…where α = 0.18 for average-key scenes (akin to automatic exposure control systems present in digital cameras [Goodnight et al (2003)]). For high-key and low-key scenes, the parameter α has to be tweaked, but an automatic estimation strategy is described by Reinhard (2003).…”

Rendering from compressed high dynamic range textures on programmable graphics hardware

“…HDR Tone Mapping Significant research has been conducted in converting a HDR image into a LDR one so that it can be displayed on LDR devices [Li et al 2005;Goodnight et al 2003;Reinhard et al 2005;Ward and Simmons 2004]. However, these techniques are mainly designed for dynamic range reduction (i.e.…”

High Dynamic Range Volume Visualization

“…In addition to reducing the range of luminance, tone reproduction can also be used to mimic perceptual qualities, resulting in an image which provokes the same responses as when viewing the scene in the real world. With the development of commodity graphics hardware, many tone mapping algorithms could be accelerated on graphics hardware [10]. Recently, a high dynamic range display device has been developed [27] based on a combination of an LCD panel and a DLP projector.…”