Tuesday, June 8, 2010
Behind The Scenes
The truth about HD video capture in DSLRs
They’re also very small sensors: the 1/3-inch sensors used in many HD camcorders have a picture area at around 4.8x3.6mm. This, in turn, means their pixels are very small (2.5 microns or less), which means their light-gathering power is low, and thus low-light and high-ISO performance is limited. DSLRs, on the other hand, have sensors ranging from APS-C (approximately 23.6x15.8mm) to “full-frame” (36x24mm), with much larger pixels: 4.3 to 8.5 microns. Video-capable DSLRs produce much better image quality in dim light and at high ISO settings, which makes it possible to shoot excellent-quality ambient-light images in situations where camcorders struggle.
Another benefit of the larger sensors in HD DSLRs is their ability to limit depth of field. You may want to concentrate the viewer’s attention on a specific subject or portion of a subject, and this is effectively done through selective focus, focusing carefully at a wide aperture to minimize depth of field so only the desired part of the subject is sharp. With the small sensors in HD camcorders, depth of field is fairly extensive, even at wide apertures, so you can’t employ selective focus as well. (Cinematographers often zoom in as much as possible, using the long end of the lens at a wide-open aperture in order to gain the shallowest depth of field, which limits composition choices to manual movement of the camera.)
Video-capable DSLRs have sensor resolutions of between 12 and 21 megapixels, which are captured at variable rates in DSLRs between 24 to 60 fps in order to produce the “moving pictures” of video. So how do DSLRs handle it? Original image data is captured at the image sensor; then the camera must enter Live View mode where the sensor produces a live data stream that’s sent to the LCD monitor. At the same time, the live data stream is written to the memory card as a compressed video file codec. In order for the image data to match the aspect ratio of HD video, it must be cropped or trimmed. Then the demosaiced image must be reduced to 1920x1080 (or 1280x720, 640x480 or at whichever video resolution you’ve chosen to shoot). Finally, once the image resolution has been reduced, the remaining data is compressed before being written to a file on the camera’s memory card.
The manufacturers wouldn’t comment to DPP one way or the other as to which downsampling process is occurring in their respective video-capable cameras, but two schools of thought have been suggested for how this occurs in a still camera: line-skipping and binning.
|One of the incredible benefits of HD DSLRs is the big range, literally, of lenses. DSLR lenses are built for precision and interchangeability, providing a healthy selection of glass with many different focal lengths and types, including fisheyes, 1:1 macros, and in Canon’s and Nikon’s case, this even includes tilt-shift lenses. All of them are available at a price point far below the expensive lenses of camcorder and film camera mounts. The majority of camcorders include a built-in, wide-range zoom, which often isn’t interchangeable. The lenses available for DSLRs are very fast, as well, and when combined with the large image sensors, they yield beautiful bokeh, which is much more desirable for most video needs than the less shallow depth of field produced by slower video lenses and smaller camcorder sensors. That being said, cinema lenses have advantages for film and video, including large-diameter focusing rings that make it much easier to precisely perform manual focus, although there are already plenty of focus-assist products available for HD DSLRs.|
Page 2 of 4