A bit confused: what’s a HEIF and why do we want 10-bit stills and video?: Digital Images Assessment


The newest cameras from most manufacturers are including 10-bit video (and sometimes stills) seize choices. The Canon EOS R5 and R6, for example, acquire each.

10-bit video, 10-bit stills modes and the flexibility to shoot ‘HEIF’ recordsdata are all options more and more being added to cameras. However what is the profit and when do you have to use these modes? We will take a look at how knowledge is captured, the way it’s saved and therefore what advantages it’s best to (and should not) anticipate from 10-bit seize.

Linear encoding

The inefficiency of linear encoding

Linear encoding devotes half the values to the brightest cease, 1 / 4 to the following cease, and so forth.

Now think about photon shot noise; the randomness of the sunshine you captured. Shot noise is actually the sq. root of the sign. So the very brilliant components of the picture (that you simply’re devoting the majority of your uncooked values to) has the best quantity of noise, as a result of the sq. root of a bigger quantity is larger than that of a smaller quantity. It does not look as noisy, as a result of noisiness relates extra carefully to the signal-to-noise ratio, not absolutely the noise degree. Nevertheless it signifies that you are capturing plenty of very tremendous element about one thing with very excessive variance.

Worse nonetheless, the human visible system is much less good at discerning element and shade in brilliant areas than it’s in darkish: you are over-encoding a loud sign that is not particularly visually significant. In brief: linear encoding is massively inefficient. (Some Uncooked compression takes benefit of this: compressing the over-encoded highlights in such a approach that has no significant impact on the picture visually or when it comes to modifying flexibility).

Not like the human visible system, cameras report mild in a linear method: twice as many photons hitting the sensor leads to a sign that is twice as giant, and is recorded utilizing a digital quantity that is twice as giant.

This implies half of your out there uncooked values are at all times consumed by the brightest cease of sunshine that you simply captured. That is simply fundamental logic: regardless of the brightest worth you captured was, half as a lot mild (ie: one cease much less mild) will probably be captured with a worth half as giant.

The upshot of that is that in Uncooked, you may solely retailer roughly the identical variety of stops of dynamic vary as your digicam’s bit depth. Or, to get trigger and impact the correct approach spherical: the analog-to-digital converter could have been chosen as a result of it has ample bit-depth to encode the indicators coming off the sensor. It is primarily a query of making certain you may seize and retain all the data coming off the sensor. Upping the bit-depth past what’s required to totally encode the sign is not going to provide you with ‘extra delicate gradation’ or ‘X million extra colours,’ it will simply imply producing a lot larger recordsdata which have recorded the noise within the shadows in additional element.

So, why can we report linear Uncooked recordsdata? As a result of it is the best factor to do from a processing perspective, retains all* of the data you first captured and is not unmanageably massive, because you’re usually solely capturing a single uncooked worth at every pixel, not separate pink, inexperienced and blue values.

Gamma encoding

Gamma encoding is the method of making use of a non-linear transformation to the linear knowledge or, extra merely: redistributing the Uncooked knowledge in a extra space-efficient method. Usually the other gamma curve is utilized while you open the picture to view it, so that you simply get again to one thing that appears like the unique factor you had been making an attempt to seize.

Gamma encoding + tone curve

As a result of this encoding is non-linear, you may squeeze some or the entire knowledge of a linear Uncooked file right into a a lot decrease bit-depth file. Nearly all fashionable cameras output 8-bit JPEGs, which generally embody round 9 stops of DR (extra when DR growth modes and adaptive tone curves are used). In precept you might in all probability match nonetheless extra, however along with the gamma encoding**, an ‘S’ curve is often utilized, which provides a pleasant punchy picture.

After gamma encoding, an ‘S’-shaped curve is utilized to most JPEG recordsdata to present a picture with a very good degree of distinction. This assigns almost 70% of the out there 256 values to the 4 stops round center gray. This reduces the scope for making adjustments.

With intelligent compression, a JPEG can simply be 1/sixth the scale of a Uncooked file however nonetheless do a very good job of conveying every little thing you noticed while you took the picture. Or, not less than, every little thing your 8-bit, Normal Dynamic Vary show can present. Nevertheless, as a result of quite a lot of knowledge has been disposed of, and since the ‘S’ curve has crushed the highlights and shadows, a JPEG gives a lot much less flexibility if you wish to make main adjustments to it.

An 8-bit file has 256 knowledge values for every color channel, and by the point you have shared these values out between 9 stops, there’s not a lot scope for adjusting the values with out gaps beginning to seem, and posterization creeping in, as an alternative of easy tonal transitions. It is an excellent endpoint, although, particularly for SDR shows.

The choice of whether or not to shoot Uncooked or JPEG is, broadly talking, a query of whether or not you propose to edit the outcomes or current them kind of as-shot. Uncooked knowledge is often 12 or 14-bit and though its linear encoding is absolutely inefficient, the very fact it hasn’t been demosaiced makes it manageable. However for many makes use of, the ultimate picture might be nicely expressed in an 8-bit JPEG. So why do we want 10-bit choices?

Log encoding – a center floor

Log encoding shares its out there values out extra evenly: most stops are given the identical variety of knowledge values, reasonably than being spectacularly weighted in direction of the highlights, as they’re with linear encoding or centered on the mid-tones like most traditional tone curves.

Log curves (Sony’s S-Log3 on this occasion), share the out there values way more evenly between the captured knowledge, retaining a very good degree of flexibility for modifying however with out the inefficiency of linear encoding. Word that the connection within the shadows shouldn’t be logarithmic (the place that method would commit extra knowledge values than the unique linear seize).

Basically, it is a intelligent approach of retaining a very good diploma of editability in a way more environment friendly file. You possibly can see why it could grow to be a well-liked approach of working with video, the place you may retain the pliability to edit, however nonetheless profit from the very environment friendly, well-optimized codecs and filetypes which were developed for video.

A transfer from 8 to 10-bit means you may have 1024 values to share out, so you may retain 4 instances extra details about every cease you captured. In flip, this implies heaps extra flexibility in case you attempt to make important changes to paint and distinction, with a lot much less danger of posterization.

Usually a producer will take a look at the efficiency of its cameras then develop a log curve that may encode many of the digicam’s usable dynamic vary. That is why most producers have ended up with a number of Log curves: you do not wish to share your 1024 values out throughout 14 stops in case your digicam’s output is unusably noisy past cease 11.

For many functions, although, 10-bit encoding of Log provides an enormous enhance in editabilty with out the file sizes changing into too unwieldy.

Why else would I want 10-bit?

So, 10-bit seize lets cameras provide way more editable video, with out the added measurement, and potential compatibility points (and authorized problems) of capturing Uncooked video. However there are different makes use of, which promise profit for each video and stills shooters.

A brand new era of TVs are actually broadly out there that may present a wider dynamic vary than older, SDR shows. An growing variety of motion pictures and TV reveals are being shot in HDR and steaming providers can ship this HDR footage to peoples’ properties.

Hybrid Log Gamma (HLG) and Perceptual Quantizer (PQ) are the 2 commonest methods of encoding HDR knowledge. Each require 10-bits of knowledge as a result of they’re making an attempt to protect a wider tonal vary than a typical 8-bit footage. Do not get fooled by the phrase ‘Log’ within the title HLG: solely a part of the curve is logarithmic. Each HLG and PQ are, like JPEGs, designed to be end-points, reasonably than intermediates.

An growing variety of cameras can shoot HLG or PQ video for playback on HDR TVs. Nevertheless it’s additionally more and more widespread for them to supply 10-bit stills based mostly round these requirements, for playback on HDR shows.

So 10-bit stills are for lifelike HDR?

What the HEIC? What are HEIF recordsdata?

The ten-bit stills modes on most up-to-date digicam shoot the single-image ‘HEIC’ type of HEIF recordsdata, that are primarily single frames of H.265 video. Regardless of being 10-bit, the extra environment friendly compression means they will retain the identical picture high quality in a a lot smaller file. Panasonic as an alternative makes use of the much less widespread ‘HSP’ format for its 10-bit stills, which is part of the HLG normal. That is even much less broadly supported than HEIF, however you are more likely to must plug your digicam right into a show to view the recordsdata correctly, both approach.

It is truthful to say that there is not a lot consistency in what the totally different digicam makers are providing. Some digicam makers solely assist you to shoot 10-bit recordsdata while you’re capturing true HDR pictures, whereas others solely provide SDR profiles in HEIF mode, and a few allow you to shoot no matter mixture you want.

From our perspective, there’s not quite a lot of level capturing 10-bit stills with typical SDR curves: the info is not saved in a fashion that is designed for modifying and Uncooked stays a extra highly effective choice anyway, so all you are doing is capturing one thing that’ll find yourself trying so much like a JPEG however is not as broadly supported.

Capturing true HDR pictures (HLG or PQ) in 10-bit makes much more sense. These can look spectacular when proven on an HDR show, with highlights that glisten in a approach that is laborious to convey in typical pictures. For the time being, although, you usually must plug your digicam right into a TV utilizing an HDMI result in view the pictures, which is not very sensible. However for us, that is the place the worth of 10-bit stills lies, photographically.

There’s quite a lot of work to be throughout the imaging trade to spice up assist for true HDR stills. We want modifying instruments to allow us to fine-tune Raws into HDR stills, simply as we’re used to doing when produce our personal JPEGs. However above all, we want wider assist and cross-compatibility in order that we will share and think about 10-bit recordsdata with out having to attach our digicam to the show. Till that is resolved, the flexibility to shoot 10-bit stills is of disappointingly restricted use.


*Or almost all in case your digicam gives lossy Uncooked compression

**Strictly talking the ‘Electro Optical Switch Operate’ (EOTF) of sRGB is not a easy 2.2 gamma transformation, however it’s comparable sufficient that it is also known as such.

We will be happy to hear your thoughts

Leave a reply

Digital Marketplace
Logo
Enable registration in settings - general
Compare items
  • Total (0)
Compare
0
Shopping cart