A recent BBC Research White Paper entitled "The Film Look: It's Not Just Jerky Motion....," written by Alan Roberts (BBC R&D White Paper WHP 053), examines the numerous attempts to make video look more like film. Many of the contributors to the look of film that must be considered when attempting to emulate it in video have been recently discussed in this column, including modulation transfer function, depth of field, and judder. If this column piques the reader's interest, the white paper itself may be downloaded at http://www.bbc.co.uk/rd/pubs/.
In Roberts' abstract, he discusses the many attempts to emulate "the film look" in video, stating that most have concentrated either on mimicking the temporal look of film or its handling of contrast, or both, but that the issues of sharpness and depth of field have rarely been addressed.
He posits that the major differences between film and video may be neatly divided into two areas: The amplitude transfer characteristic, which deals with contrast range, and the modulation transfer function, which deals with "everything else."
MINIMAL DIFFERENCE
The first point Roberts makes is that, despite the commonly held belief, film can handle a contrast range of at least 10 f-stops-or 1024:1-while video can only handle about 5 stops or 32:1. There is, in fact, not a great deal of difference between film and properly setup HD cameras in this respect. He says that negative film has a central exposure range of about two decades over which the exposure versus density curves are linear with a slope of about 0.9. Outside the linear range, the curves compress at the two extremes, covering a total of about four decades and forming the familiar "lazy S" pattern. Thus film delivers a linear contrast range of about 100:1 or 6.5 stops, and a further 5:1 or two stops at either end of the contrast range-which are significantly crushed-for a total range of 2000:1 or 11 stops. He then does an extensive analysis of HD cameras that leads to the conclusion that, if the controls are properly set, they are capable of about the same 11 stops of contrast range.
He then looks at modulation transfer function and modulation transfer factor, explaining that modulation transfer factor (designated mtf) is a system's gain at a specified frequency, while modulation transfer function (MTF) is the distribution curve of mtf over a range of spatial frequencies. MTF is important because it is responsible for the detail seen in a picture, which affects "everything else." In addition to the general sharpness of images, it provides the visual cues for object placement and image location, and thus for the portrayal of motion. This causes the list of things affected by MTF to include the perception of judder. It also affects depth of field. One of his central conclusions is that if video is to mimic film, it must mimic film's MTF reasonably well.
This paper, written in a 50 Hz milieu, discusses the kind of judder characteristics of 50 Hz systems, which for 24fps film (speeded up to 25fps), is the straightforward result of each picture being shown twice. We know that the result of showing each picture twice is a distortion of the timeline such that in one of the showings, a moving object will be displayed in the wrong place or at the wrong time. The perceptible effect varies from a vague "restlessness" in the pictures when the spatial displacement between showings is small (motion is slow), to the extreme, when the spatial displacement between showings is large (motion is fast), which causes the visual system to separate the two showings of the moving object into two motion paths, resulting in a single object being seen as an adjacent pair of objects. This kind of judder may be seen in any movie theater, where 24fps pictures are double-shuttered. In the 60 Hz television world, 24fps are displayed with 3/2 pulldown, and the resulting asymmetry between adjacent pictures makes the judder's appearance more complex, but the principles are the same. Judder is an important factor in the "film look," and the degree of judder perception depends on the sharpness of the images, which is controlled by MTF.
The author describes the important role depth of field plays in the subjective appearance of material shot on film, citing the often-heard complaint that the depth of field of video is too great. This makes selective focus-a very important creative tool used to put the primary subject in focus while the background is defocused-very difficult to achieve.
You might recall that the depth of field is the range of object distances from the camera lens over which everything appears to be in focus. As lens aperture or f number decreases, depth of field increases, and, if the f number is held constant, as the size of the image decreases, depth of field also increases. The result is that for the same f number, a 2/3-inch CCD camera will have about 2.2 times the depth of field that a 35mm camera will have. As many 35mm movie lenses are designed to function optimally at maximum aperture, which may be as much as f/1.6, there is no chance that a 2/3-inch video camera will achieve the same depth of field as a movie camera. So while it is technically possible to use a 35mm movie lens with a 2/3-inch video camera, the angle of view and depth of field will not mimic 35mm film at all. One solution is to mount the 35mm lens so that its image is generated on a ground-glass screen outside the camera (increasing the image size to that of a 35mm film frame), and using relay optics to route the image to the CCD sensors. This would increase the size, complexity and price of the video camera system.
NOT 'WELL RED'
It may not be immediately apparent that MTF affects skin-tone resolution. Film is made with its red-sensitive emulsion layer on the bottom, next to the base, and the blue-sensitive layer on top, with the green-sensitive layer between them. Because red light has to travel through the other two layers before reaching the red-sensitive layer, film is inherently less sensitive to red than to other colors-in colloquial parlance, film is said to "not do red well." A result of this lowered sensitivity to red is that detail in skin tones is muted, visually softening them. Video does not exhibit this reduced response to red, and this is probably a principal reason for the frequent complaint that video makes skin tones look harsh.
Roberts concludes that it is now possible to operate HD cameras so that film performance is convincingly imitated by presetting the camera in certain ways:
• To mimic film motion judder, shoot in progressive scan mode (24p here in the U.S.), with the shutter set at about 50 percent to imitate the 180-degree shuttering of the film camera.
• Preset the gamma corrector, black stretch and knee to capture an 11-stop contrast range.
• Video cameras cannot match film MTF, but parameters such as aperture correction and detail enhancement may be set to bring them closer to filmlike MTF.
• Video cameras have far more depth of field than 35mm movie cameras, but there are some ingenious ways being developed to mitigate this effect when 35mm lenses are used on video cameras.
• Video has better color resolution in red than does film, and this makes skin tones look harsh. Some recent cameras have a feature that detects and softens skin tone. This feature should only be adjusted in the field using a high-resolution monitoring system to determine the effect, as it is very easy to overdo it.
This BBC white paper on achieving the "film look," using HD video cameras is one of the better treatments of this subject-I highly recommend it.
