Linear perspective

One-point linear perspective: Parallel lines converging towards a single vanishing point on the horizon.

Linear perspective (also referred as diminishing perspective) is both a mathematical theory of projecting 3D spaces on a 2D plane and a related technique of depicting space in a drawing. Incidentally, this is exactly what our photo image plane does with the photographed space.

But here we are interested in perspective in the context of perception. That is, the way objects appear to the eye depending on their distance. All perspective cues exploited by the brain follow from the simple fact that objects appear smaller with increased distance from the eye. When the absolute size of an object is familiar, its distance can be judged depending on the size of its projection on the eye’s retina. When the absolute size of an object is unknown, but there are at least two objects of the same kind in view, the relative size of the objects suggests a notion for their relative distance. The property of parallel lines to converge towards the horizon is also a good depth cue.

Lenses and perspective

Any rectilinear lens creates a (linearly) perspective image of the scene (a fisheye lens renders curvilinear perspective). A common misconception has it that wide lenses strengthen perspective and long lenses weaken perspective. This is not true. Perspective is entirely dependent on viewpoint. Relative positions and relative sizes of objects only change if the eye moves. It is not the angle of view of the lens that manipulates perspective, rather the shift of camera viewpoint forward with a wide lens, and backward with a long lens, in order to frame a subject in a similar way. This change of camera viewpoint creates the perspective differences often wrongly associated with the angle of view of the lens.

With the same size of a reference object in the frame, the wide lens exaggerates foreground-background relations and space appears expanded. Conversely, the long lens seemingly compresses and flattens space. Consequently, one good general rule for consistent space representation is to avoid mixing focal lengths when preserving the size of an important object in the frame. So this rule does not apply when changing from a wide shot to a close-up, etc, because objects change their sizes in the frame. The Vertigo effect (also known as a dolly zoom shot) demonstrates what happens when the change of focal length while preserving subject size is realized in a single continuous shot – space seemingly expands (pushing the camera in while zooming the lens out) or contracts (pulling the camera out while zooming the lens in).

In this context, lens choice is an instrument for perspective control. A normal lens presents a certain naturalism in space rendering. This is the focal length of choice when the camera must disappear. What exactly is a “normal lens” is worth an article of its own, what with all the misconceptions surrounding the term, especially in the moving pictures realm. Quick definitions for the purposes of this article: a “normal” image has its center of perspective at the viewer’s eye; wide-angle images have their center of perspective in front of the viewer; telephoto images have their center of perspective behind the viewer. Any non-coincidence of the viewing position and the center of perspective in the image creates a perceived perspective distortion, which can be used for artistic purposes.

Children of Men was mostly shot with an 18 mm lens (on Super 35).

Wide-angle lenses can render a stylized dramatic space and exaggerate movement on the depth axis; the intimate viewpoint adds a feeling of immediacy. Films by Stanley Kubrick, Terry Gilliam, Jean-Pierre Jeunet and Barry Sonnenfeld often exploit these characteristics. Telephoto lenses can emphasize the graphic qualities of the image and promote the abstraction of a plane from the space; the viewpoint is detached and formal (or voyeuristic, depending on context). Kurosawa (who was very particular about these things) used mostly long lenses together with multi-camera setups during the second half of his career.

Tinker Tailor Soldier Spy used long lenses and distant viewpoints extensively on exteriors to build a feeling of spying on the action.

Forced perspective

Forced perspective is a technique that exploits the principles of familiar size and relative size, and plays with the expectations of the brain to create an illusion. Depending on context, objects are made to appear either closer or further away than they are, or smaller or larger than their actual size. In the first case, cheating the size of the object manipulates its perceived position. In the second, cheating the position of the object manipulates its perceived size. Size manipulation was perfected in the Lord of the Rings trilogy. Tweaking actors’ positions and moving set elements around in sync with the camera allowed keeping the illusion even on moving shots.

Small props and small people in the background fake a longer train and a deeper scene. Hitchcock shot this on a stage.

But the topic of this article is more concerned with faking distances. In its most popular form the technique uses objects smaller than their real life size to increase the perceived distance. This is most often done to effectively increase the perceived depth of the entire set. Filmmakers adopted this practical approach of faking depth very early. It was already an art in the days of the German expressionists. Using mattes to fake distant backgrounds is nothing more than a special (but simple) case of forced perspective. These used to be painted panes strategically placed on the set. Nowadays this is done mostly in post with greenscreens and CGI.

Aerial perspective and color perspective

Aerial perspective is usually demonstrated through mountain vistas or cityscapes. I am staying true to tradition.

Aerial perspective describes the effect of atmosphere on scene appearance. It is often called atmospheric perspective or tonal perspective, the latter term widely used in the context of visual arts. The scattering of skylight from various air particles creates a veil of sorts over the scene. Effectively, some amount of scattered skylight is added to the reflected scene light. The longer the distance from the viewer to the object, the stronger the effect. As a result, distant depth planes have lower contrast, lower saturation and higher brightness than closer planes, and their tones appear to converge towards the luminance (and color) of the distant sky. Aerial perspective is another reason for the presence of texture gradient in large-scale scenes. The eye needs contrast to differentiate detail, so there is a gradual loss of fine detail perception with increasing distance. And since reflected scene light is also scattered by the atmosphere, object definition itself is affected in the first place: atmosphere acts as diffusion.

The accumulation of scattered skylight also creates a color shift. Since short wavelength light scatters more, there is more green and especially blue in the scattered daylight. That’s why distant objects seem to acquire a blue cast. Consequently, the conditioned brain readily agrees that warm tones in an image tend to advance, and cold tones tend to recede. This is sometimes called color perspective.

Smoke helps define depth planes through contrast differentiation. Top: Selective lighting and a bit of smoke help lift the midground from the foreground and the background. Bottom: Smoke is also a way to pop silhouettes and visualize light shafts.

Atmospheric conditions are certainly hard to control, but the tonal perspective effect can be mimicked on a smaller scale, and that includes interiors. This is achieved through the use of artificial smoke and fog. Their higher particle density leads to much more pronounced atmospheric effects and faster color modulation with increasing distance. So infusing atmosphere into the set also creates atmospheric perspective. Even a little bit of smoke adds some fill to the shadows in the far end of the set, establishing an axial gradient of contrast. This lowers the contrast of the background and softens it, helping the subject to stand out. Rain can work in a similar fashion as a device for enhancing depth. The water droplets are much bigger particles, but they also obscure objects, scatter light and veil the distance.

You can read the the first part of the Creating Depth series here. And the next part is on light and depth.

Creating Depth, Part 2: Perspective

Usability requirements for lenses for video

Any lens can be used to shoot video, as long as you can mount it on the camera. That said, there are a couple of features, which make the job easier.

Smooth aperture is the most underestimated lens feature by new videographers. Image consistency generally needs fixed sensitivity of the capturing medium (fixed ISO) and fixed shutter speed. This leaves aperture as the only mean for precise exposure control. And precise control you will need, once you have to (near) perfectly match the exposure of a couple of lenses between shots in a sequence. Well calibrated t-stop markings allow for perfect exposure matching. In fact, t-stops are critical when shooting film. With digital one can get a perfect exposure match even without t-stop markings, as long as there is smooth aperture control available. This is achieved with through-the-lens exposure tools like a waveform monitor or a spotmeter and a solid tonal reference (usually a grey card). Exposure matching during the shoot is not mandatory, but saves matching time in post.

Next is the tight and smooth focus ring. A good focusing ring has decent travel for focus precision; has no play; ideally, has a hard stop at infinity; either allows for a focus gear to be attached, or is standard pitch geared itself for use with a follow focus.

Lens image rendition in relation to video

Lens resolution is largely overrated for low-end video. Sensor line skipping (used for video in most photo cameras) is brutal in decimating image definition and introducing fake detail from aliasing. Pixel binning is much more forgiving, essentially being localized downsampling. Image quality is then further degraded by video compression. Any decent lens will do fine in terms of sharpness for DSLR video. On the other hand, lens resolution matters a lot when high quality 4K video is concerned, or when recording RAW video from a small sensor (as with Blackmagic’s camera).

Lens contrast needs more attention. First, higher contrast images will appear to have more detail. Second, higher contrast images will endure somewhat better through heavy video compression due to their better value differentiation. Low contrast lenses, on the other hand, can slightly boost the captured scene contrast. Lens contrast is related to coatings: low contrast lenses are usually uncoated or single-coated, so they will also flare more.

Color rendition varies a lot. Older lenses often render warmer (due to benefits for black and white film). It is best to have neutral color rendition, but lens color can be modified through white balance camera options, or in post.

Out-of-focus rendering (bokeh) is largely unaffected by video specifics as its appearance is defined over (relatively) large areas of the image. So it is valid to take it into consideration when choosing lenses for video. Same with lens distortion and vignetting. Using full frame lenses on crop bodies pretty much eliminates vignetting problems though. Also, some cameras will correct these defects for native lenses. Abundant chromatic aberrations will be visible in video.

Video lenses vs. still lenses

Samyang cine lenses have smooth aperture, t-stops, and geared focus and aperture rings

Unsurprisingly, video/cine lenses are better suited for shooting video than photo lenses. As a minimum, video lenses have smooth aperture control and good focus rings with standard pitch teeth for use with a follow focus. They usually feature t-stop markings instead of f-stop markings on the aperture ring. The distance scale is often to the side to make it visible to a focus puller, detailed, and precisely calibrated. Same with depth of field scales, where available. Photo lenses will mostly have the distance scale on top, so that the photographer can have a quick glance at it at will. High quality video lens sets like the Carl Zeiss compact primes are also color matched and feature matching front diameters (for use with matte boxes). Overall, image characteristics are matched across the set, including generally matching out-of-focus rendering due to the consistent number of aperture blades.

Video and cine lenses have a flaw though: they are usually quite expensive. Only recently Korean and Chinese lens manufacturers are getting into the market, with Samyang (also rebranded as Rokinon or Bower in the USA) and SLRMagic rehousing their photo lenses for video use and keeping the prices down. But in general, still photo lenses, especially some legacy 35mm lenses, are significantly cheaper.

Native mount lenses vs. adapted lenses

If you are using a photo camera for video (HDDSLR or a video capable mirrorless), you are likely invested in native glass already. Modern photo lenses, except for some third party lenses, are universally autofocus and usually feature electronic aperture control. Autofocus is still largely irrelevant for video, so the lenses are mostly used in manual focus mode. Autofocus lenses have a couple of side effects though. They usually have relatively short focus throw (for faster AF), which may limit manual focus precision. And their focus ring rotation don’t stop at infinity when in manual focus mode, but rather goes beyond infinity. That’s because the autofocus system usually needs to go beyond the exact focus point and then back a little when hunting for focus in order to get the sharpest result.

Electronic aperture control can be either a positive or a negative. On some cameras (Nikon D4 or Canon C300, for example) aperture can be adjusted in 1/8 stop steps, allowing some quite precise exposure control. Typically though, on photo cameras the aperture is controlled in 1/3 stop steps. This is still better than most manual focus and manual aperture legacy lenses which allow aperture control in either half stops or full stops. But manual aperture can usually be declicked, resulting in smooth aperture rings. Declicked aperture means both precise exposure control (with the help of through-the-lens exposure tools) and the possibility for basic aperture pulls during shots. The latter can’t be done with stepped electronic aperture.

Using lenses recognized by the camera body (this generally means lenses from the camera manufacturer) has additional benefits. In-camera processing can fix some optical defects. For example, Canon DSLR cameras have an option for peripheral illumination correction (vignetting correction) available for Canon lenses. Some cameras (Panasonic, Samsung, etc) will also optionally fix distortion on lenses they recognize.

Then there is image stabilization. This feature of some native mount lenses is very beneficial for handheld video, more so with long focal lengths. If handheld is your thing, it is probably wise to stick with native lenses with IS.

So what do legacy manual focus lenses have to offer?

The focus path of the Carl Zeiss Jena Flektogon 35/2.4 is more than 250 degrees.

Quality old manual lenses usually have tight and smooth focusing rings, which are vastly superior to AF lenses. The focus throw is often noticeably longer than AF lenses. As mentioned above, manual aperture rings (where available) can usually be declicked for smooth aperture control. Legacy lenses sometimes have “character” (which is glorified optical defects). Lenses from the 60′s and earlier will generally be uncoated or single-coated and will flare easily. This can be used for lowering contrast or for artistic effects. Legacy lenses often have sturdy metal bodies, unlike the plastic bodies common with modern lenses.

But the main attraction is price. With patient shopping, you can often collect a whole set for the price of a quality modern lens. That is, if you don’t jump on the Leica train. Relatively cheap manual lens systems include Super Takumar (M42), Olympus OM, Canon FD, Carl Zeiss Jena (M42), Pentacon (M42), Pentax K, Yashica ML (Contax/Yashica mount), Minolta MD, to name a few. Going up in price, there are manual Nikon F, Contax Carl Zeiss and Leica R.

Adapting lenses

Olympus OM to Canon EF adaptor. FFD of the OM system is 46mm and FFD of Canon EF is 44mm, so OM lenses can be adapted to Canon EOS cameras.

The main factor that decides if a lens can be adapted to a camera mount is flange focal distance (FFD). This is the distance from the camera mount to the sensor plane. In general, lenses from a system with longer flange focal distance can be adapted to a system with shorter flange focal distance. A lens with a shorter FFD will need an adaptor with optical elements in order to be able to focus to infinity (if possible at all) on systems with longer FFDs. It is best to not bother with these. The common way to adapt is through a separate adaptor between the lens and the camera mount. But when the difference in FFD between the two systems is small, lenses can often be adapted through replacement lens mounts that take the place of the original lens mount. The replacement mount is manufactured to more or less equalize the FFD. Popular replacement mount routes are Canon FD to Canon EF, M42 to Nikon F, Contax/Yashica to Sony Alpha, Leica R to Nikon F, to name a few. The following table lists the flange focal distance of some popular camera mounts for reference.

Mirrorless mounts are the easiest to adapt lenses to. You can put pretty much any lens ever made on a Sony E-mount camera. Among (video) DSLRs Canon EF cameras are the easiest to adapt to, and Nikon the hardest. Due to mount specifics Sony Alpha cameras often require replacement mounts to be installed on the lens even though the FFD is only 0.6mm longer than Canon’s.

FFD is only one part of the equation though. Some lenses simply can’t fit physically on a camera even if they have a longer flange focal distance: they could be too wide, or they could hit the mirror (in the case of a SLR camera). For example, Arri PL lenses have wide mounts and protrude deep. Adapting can be a tricky business. For example, some Leica R lenses work on APS-C Canon DSLR cameras, but hit the mirror on full frame Canon DSLRs. So before investing in a lens make sure it can be used on your camera. Another issue is adaptor thickness. In order to ensure infinity focus adaptor manufacturers routinely make adaptors a hair too thin. This results in unreliable distance scales and annoying loss of the hard stop of the focus ring at infinity (where such a hard stop was previously available). Depending on adaptor construction, adaptors can sometimes be shimmed to the precise thickness through trial and error.

Prime lenses vs. zoom lenses

Primes’ most important advantage over zooms is faster maximum apertures. This means better low light capabilities and the option of shallower depth of field. Their generally better sharpness only matters when good quality video compression and/or high resolution are available. Think dedicated video sensors, RAW video or 4K. Primes are usually smaller and with less glass elements. This translates to lighter rigs.

Non-cine zoom lenses often extend forward a lot making matte box operation hard

Zooms, on the other hand, are all about flexibility and convenience. A single lens covers a whole range of focal lengths. This means less lens changes and lugging fewer lenses around. There is also the added bonus of color matched focal lengths over the zoom range. With parfocal constant aperture zoom lenses you can also zoom during a shot. So make sure your zoom lens is parfocal and not varifocal, if you plan to do this. Note that constant aperture is not always constant. There are often variations in true aperture over the zoom range. This may lead to slight exposure changes during zooms. Push-pull zoom lenses will need readjustment of the follow focus (if you use one) as the focus ring moves forward or backward with zooming. Modern zooms are mostly of the two-touch type, with separate zoom and focus rings. Photo zoom lenses (unlike cine zoom lenses) often extend a lot when zooming, which may either prevent their use with a matte box, or at least require matte box readjustment after changing the focal length. Then again, one can do fine without a matte box, especially when lens changes are rare. Photo zooms rarely have apertures wider than f2.8. If you need low light, you are better suited with primes.

Some thoughts on lens set building

If you’ve never shot seriously before, it is probably best to start with a standard zoom. Get familiar with the focal lengths, find out what you like and proceed from there. You can shoot a feature with a zoom lens. You can shoot a feature with a single prime. Buying a lens only because someone raves about it on the Internet is a bad idea. Get only what you need, when you need it. Build up gradually. It is usually best to build a set of lenses from the same system – only Nikon lenses, or only Canon lenses, or only Leica R lenses, etc. Some lens systems are better matched than others, but it is best to aim for serial numbers close to one another. Mixing lenses is the easiest way to get in trouble with matching color, contrast, or just “feel” between shots. But you should not absolutize this. Content is more important than the occasional mismatch, and sometimes the mismatch can even be used to an advantage. And sampling various lens systems will also give you a taste of what is out there. There is more on lens color in the article on color matching lenses for video. Matching the filter thread diameter of the lenses will let you work without step-up rings, both with and without a matte box. Luckily, lens manufacturers often utilize only one or two filter sizes within a single lens system.

Random tidbits

• Nikkors’ focus rings rotate in the opposite direction compared to pretty much any other lens.
• You may want to consider the minimum focus distance when choosing a lens. A lens with short MFD can give you near-macro shots without the need for a dedicated macro lens.
• Some very small lenses have the focus ring so close to the camera body that using a follow focus is hard or impossible.
• Some lens repair shops do cine mods to still lenses. These usually include installing focus gears, common fronts, and declicking the aperture.
• flickr is your best friend when you need to research how a lens renders.

Choosing Lenses for Video

Carl Zeiss compact primes cp.2 are currently the cheapest truly matched cinema lenses

Unfortunately for most video enthusiasts, cine lenses are both quite expensive and usually manufactured to fit cinema mounts like Arri PL or Panavision PV. Only recently lens sets like the Zeiss compact primes have started to appear for hybrid and photo mounts. Based on the latest generation of Zeiss SLR photo lenses, the Zeiss compact primes have been reworked to offer a significant degree of consistency, plus interchangeable mounts. But while not as expensive as bigger cinema lenses, Zeiss compact primes are not exactly cheap.

This leaves most large sensor video shooters in the photographic lenses camp. Photographic lenses present a lot of challenges for video, but they have a significant advantage: price. Selecting photo lenses for video is an art in itself (and possibly a topic for another article), but here we will focus on lens color and color matching. First, a quick word on a related subject.

White balance and lens color

One advantage of digital cameras over film is the ability to easily tweak white balance. Color film stocks are balanced for some specific color temperature, usually 5500K for daylight and 3200K for tungsten light. Color balance can be adjusted further during post-production, either chemically, through printer lights manipulation, or in DI.

Digital sensors are also optimized for specific spectral response (usually biased towards daylight). But by applying gain on individual color channels the signal can be white balanced in-camera to pretty much any desired color temperature. Furthermore, while film is usually only balanced on the orange-blue axis and expects green-magenta neutrality, digital video can also be white balanced on the green-magenta axis. For RAW video the white balancing decisions can be deferred to post. For video transformed to some working color space for recording the white balanced is baked during RAW conversion in-camera.

In-camera white balance means that lens color is less critical for digital video compared to film. The camera can be manually white balanced with a grey card after every lens change and this through-the-lens balancing will lead to neutral rendering. And this is how a lot of videographers work. But this practice deprives the videographer from a very nice tool: white balance can’t be used for creative purposes. Typical creative white balance uses include colder than neutral balance for winter or overcast feel and warmer than neutral balance for night indoor scenes. Of course, this can be done in post. But too much post-processing is bad for low precision video. Color matched lens sets, on the other hand, allow such creative choices to be dialed in when the sequence starts as subsequent lens changes do not introduce color deviations.

Photo lenses and color

Ideally, a lens should be completely neutral in color rendering. In reality it is not quite so. Lens color rendering is dependent on coatings and glass: both can cause tints. But if two lenses use the same glass and coatings, they will almost certainly render color in a very similar way. “Almost”, because tinted glass or coatings will generally lead to heavier tints (tint stacking) in lenses constructed from more glass elements.

Early coated lenses often exhibit yellowish tints. One reason is that warm coatings render skin lighter and skies darker in black and white, resulting in a favorable tonal separation. Earlier Leica lenses are like this. Some lenses from the 50′s to the 70′s used thorium dioxide to increase the refractive index of the glass. Thorium radioactivity leads to brownish tints over time. Well-known examples include some Pentax Super Takumar and various Kodak lenses.

The wide adoption of color film led to manufacturers developing coatings with a more neutral color rendition. Nevertheless, color varies not only among lenses from different manufacturers, but also in a series of lenses from the same manufacturer. It is common sense, though, that lenses produced by the same manufacturer close in time have higher chances of being made from the same glass and with the same coatings, seeing as both of these usually don’t change very often. The Zeiss Contax/Yashica mount lenses are good examples of consistent color in lenses stretching production over a significant period of time.

The following image shows a grey card rendered by various lenses. The white balance is set for the first lens and left as is for the others. Consequently, any tints manifested are relative to the first lens. Note how the Leicas are very close to each other (they were manufactured in the same year). Same with the two Contax Zeiss lenses (an early AEJ Planar and a somewhat later MMJ Distagon). The Leicas are noticeably colder than the Contax Zeiss lenses (a difference of around 20 mireds). They also have a green tint.

From left to right: Leica R Summilux 50/1.4, Leica R Summilux 80/1.4, Carl Zeiss Contax Planar 50/1.7, Carl Zeiss Contax Distagon 28/2.8, Carl Zeiss Jena Flektogon 35/2.4

Color matching photo lenses in camera

Differences in color rendition between lenses can be easily quantified. One method appropriate for DSLR cameras takes advantage of their ability to shoot RAW still images. This involves shooting a RAW image of a grey card under a consistent light. Manually white balancing the image shot with each lens in Lightroom, Camera Raw or a similar RAW development software and comparing the resulting color temperature and magenta/green tint values will then reveal the relative differences in color rendition. Note that color temperature needs to be converted to mireds to yield a translatable result that can be used for color matching.

Canon DSLR cameras can be WB corrected independently of white balance choice

Some digital cameras have options for white balance shifts (or correction). For example, on a Canon DSLR camera white balance can be corrected from -45 to +45 mireds on an amber-blue axis in 5-mireds steps. There is also a similar -9 to +9 step correction on the magenta-green axis. While originally intended for correction under non-full spectrum light sources exhibiting color spikes, these options can also be used for in-camera lens color matching. For example, the Zeiss Contax/Yashica lenses from the above test can be decently color matched to the Leica R lenses by ticking the WB shift 4 steps towards blue and 4 steps towards green.

The Magic Lantern firmware for Canon DSLR cameras offers another way to measure color differences. Its point-and-white-balance feature allows the white balance to be set by pointing the camera to a neutral surface (ideally, a grey card). The calculated color temperature and WB shifts are immediately displayed for reference. This method has the advantage of yielding information directly in camera specific white balance terms. For better precision you should do this test under daylight because Canon DSLRs only set WB color temperature in 100K increments.

Once the relative differences in color rendition are measured against a known base setting, creative manual white balance can be used even with photo lenses. White balance can be preset for specific effects and left untouched for the whole sequence. Any lens change will only require dialing the respective WB shift values. Note that this will only help match color. Differences in contrast or out-of-focus rendering will still be present if they did exist in the first place.

Color Matching Lenses for Video