![]() ![]() In the highlights, the situation is a bit more complicated than it can be supposed, because in vast majority of cases only the linear portion of the characteristic curve of digital image sensor is used in consumer cameras (there are a few exceptions to this rule, like some Panasonic cameras at lowest "actual" ISO setting). There's nothing unexpected here - first the resolution falls gradually, then quickly. -9.6 EV - in practical terms, it's fallen to zero - only the very large, 100-pixel objects are somewhat visible. ![]() -8 EV - the resolution falls by a factor of four.-6 EV - the resolution falls by a factor of two from optimal exposure.-4 EV below middle gray - the resolution has fallen by about 20% (at the shot whee EV = 0 the dashes were read around the 10 scale mark region, for EV = -4, around the 8 scale mark).As it is easy to see, when exposure decreases (this can be the shadows on a normally-exposed shot, or middle gray on an underexposed one), the resolution falls gradually and steadily:.Here are four shots (to see them full-sized, please click) If you take a picture of this target, bracketing the exposure using the shutter speed (at the same aperture and ISO speed setting) and then (within reasonable limits) compensate for exposure ("pull the shadows back into middle gray") then we will see the following: Shadows For this and all following examples, the RAW has been rendered in 1:1 scale, shots were taken with a Zeiss Makro Planar 100/2 lens set to f/8, the target was in the center of the shot at a distance of 3.5 meters, with the entire shot of the target fitting into a 1000x1000 pixels square.EV = 0 - the exposure that is used is such that the gray background of the target in the RAW was at 12.7% of the clipping level: 3 stops below clipping is the current generally-accepted place for middle gray in RAW, coming from the idea that 2.45 stops (18%) is obviously too little in practice, 4 is a bit too much, and we'd still like to have a nicely rounded value (to get exactly -3 EV, 12.5% should be used, log2(0.125) = -3 however 0.5 EV down from 18% is 12.7%, making it easier to compensate the exposure when a standard 18% card is used).Here are two resolution targets that differ by a factor of 2 in size, since during the preparation stage it was unclear which size would be optimal (from now on, we will be using the larger, upper target illustration).and with details of known size was photographed (12pt, 18pt, 30pt font).A target with known contrast (gray text on a gray background with contrasts of 0.5, 1, 1.5 EV).What we used previously (Old Approach)įour years ago, while studying a newly-bought Canon EOS 6D, we tried to approach the program of practical dynamic range in the following way: Depending on the quality demanded of the image (which will depend on presentation size, viewing distance, other viewing conditions such as screen resolution, etc.,), the practical dynamic range for the specific camera will be different, even for the same ISO setting. It's just that with the lowering of exposure, small details disappear, the contrast between bigger details diminishes, color fidelity becomes worse. In reality, any practicing photographer knows that the degradation of the image in the shadows (or because of low exposure) happens gradually, there is no strict demarcating line. In part, if there is noise reduction or bit depth reduction taking place in the camera, then the level of noise will be (relatively) low, but that very same noise reduction will destroy small low-contrast (and sometimes not so low) details. Furthermore, measuring DR by the noise in the shadows doesn't account for the possible processing of the signal in the camera. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |