Shuttering methods and the artifacts they produce

Abstract

When exposure times were measured in minutes, the opening and closing of the shutter was essentially instantaneous. As more sensitive films and brighter optics became available, exposure times decreased, the travel time of the shutter mechanism became increasingly significant, and artifacts became visible. Perhaps the best-known shutter artifacts are the spatio-temporal distortions associated with photographing moving subjects using a focal-plane shutter or sequential electronic sampling of pixels (electronic rolling shutter). However, the shutter mechanism also can cause banding with flickering light sources and strange artifacts in out-of-focus regions (bokeh); it can even impact resolution. This paper experimentally evaluates and discusses the artifacts caused by leaf, focal plane, electronic first curtain, and fully electronic sequential-readout shuttering. Introduction The capture of a properly exposed image requires balancing of the various exposure parameters. Sensitivity to changes in exposure factors in general is logarithmic, so APEX (Additive System of Photographic Exposure) encodes all parameters as log values such that doubling or halving the parameter is encoded by adding or subtracting one from the APEX value of that parameter. The result is that equivalent exposures can be determined by the simple linear equation: Ev = Bv + Sv = Tv + Av The exposure value, Ev, represents the total amount of image-forming light. In other words, two exposures are expected to produce “equivalent” images as long as Ev is the same. The values of Bv and Sv are essentially constants for a given scene and camera. The metered luminance of the scene being photographed is the brightness value, Bv. The speed value, Sv, represents the light sensitivity of the film or sensor – the ISO. In digital cameras, the value of Sv typically is determined by the combination of quantum efficiency, analog gain, and digital gain. However, the quantum efficiency is not easily changed after manufacture, so manipulating the analog and/or digital gain to increase the ISO effectively reduces dynamic range. The remaining parameters, Tv and Av, are the things that can be directly controlled by the camera for each capture. The time value, Tv, represents the exposure integration period, commonly known as shutter speed even for systems that lack a mechanical shutter. This is the key parameter of concern in the current work. More precisely, the current work centers on characterizing the subtle differences caused by various implementations of shuttering. For example, some shuttering methods give all pixels the same duration of exposure, but do not expose Figure 1. Still image from high speed video of leaf shutter all pixels during the same time interval – thus causing specific types of artifacts. The aperture value, Av, represents the rate of light transmission through the lens. Using a perfect lens, Av is determined solely by the aperture f /number, which is simply the ratio of the lens focal length divided by the diameter of its circular aperture. However, for real lenses, reflections and other imperfections reduce the light transmitted by a small amount, so it would be more correct to say that Av is determined by the transmission-corrected effective f /number, or T/number. The size of the aperture is typically adjustable either using an iris or by inserting a Waterhouse stop, and would seem to be unaffected by the method used to implement Tv. However, as the current work shows, the effective aperture size and shape can be changed dynamically during exposure depending on how shuttering is implemented. The goal of the current work is to experimentally evaluate and discuss how the method for implementing Tv – shuttering – produces artifacts in the captured image.