Hyper on Dynamic Range

Even photography itself, in the days of degeurrotype, was a novelty and a little bit of a gimmick.

Many of the gimmicks and fawcets of cameras, films, lenses, filters and now sensors in their path of evolution have come to roost as established styles and techniques.

Probably what is the latest of the emperors-new-clothes to many luddites, is Hyper Dyanamic Range in post processing and the latest enhanced, broad dynamic range nascent in-camera, surpasing that of the human eye even. HDR is here to stay.

DR- dynamic range a measure of the extremities of light intensity which can be captured on the sensor and subsequent image file for the given exposure. This means that the nuances of light and dark extend so far as becoming pure white, or pure black shadow at a given upper and lower level of light intensity.

What is actually equally important to image quality, is tonal depth. Despite modern digital cameras proporting to have the same technical tonal depth in number of bits, different cameras render these better in RAW and in-body jpeg. Combining a wider DR with excellent TD means that more detail and nuances of colour and brightness can be captured.

How does this add up to the pretentious and contentious "IQ" ? Well you can argue a lot about subjectivity in this, but if you are a 10 - 12 mpx Four Thirds / mFT owner then you know all about limited dynamic range. Blown highlights where large blocks of white make a very intrusive presence on photos from scenes with high contrast taken on an average or subject spot exposure.

High or broad DR means you are capturing more tonal depth at the extremes of the intensity of a scene. This can now surpass the ability of the human eye, leading to a slightly surreal or jamais vu feel to especially dawn and dusk images. The additional nuances allow us to see more detail and more texture. This means the eye, sorry brain, picks up on additional qeues for depth and interest in a shot, and basically makes the perception of realism. Alternatively the surreal or the presque vu or jamais vu come through.

Okay now I am getting a little arty and subjective about how good DR and corresponding TD make for a good shot. Essentially good DR also means that more of the scene can be reported as being properly exposed or at least "pleasantly". When combining several images of different exposures as layers in post processing software, the very best of each can be employed in a final image.

Technically you can measure DR and tonal depth as finite, and DR from the various tests is often worth looking up for a camera you want to buy, and comparing to in particular the excellent DR of the Nikon D7000 and the Nikon D3S and then to rival models vying for your hard earned wonga.


But what about doing your own HDR on a four thirds camera? We want to avoid both blow highlights AND heavy shadows with in my case, the Olympus / OEM CMOS "red spotted dabs" syndrome in shadows and highlights.

Here is my next experimental path in The GIMP, starting asap, as expressed by JPMATH on flickr.

jpmatth (70 months ago | reply)


here's the entire process, simplified where possible:

* set the 20D's auto exposure bracketing to 1 stop.
* stand really still, and squeeze the shutter until three shots are taken (that would be 0, +1 stop and -1 stop). try not to move the camera (didn't have a tripod with me).
* developed each RAW file with the same settings (with recipes in DPP), and exported to highest quality jpeg (to retain the EXIF data).
* opened the three jpegs in gimp as three separate layers, with the 0 exposure at bottom, then -1 on top of that, then +1 at the top of the stack.
* moved the layers around until each element of the scene lined up as much as possible (because i was handholding, each exposure was a few pixels off from the others even though they were taken milliseconds apart)
* added transparent (black) layer masks to all but the bottom layer
* painted white on the layer masks where i wanted sections of each layer to show through.
* flattened everything and cropped the edges where the misalignment was obvious.
* a little unsharp masking to hide the slight softness of the misalignment
* saved to a final jpeg with quality at 90, no smoothing, subsampling at 1x1,1x1,1x1 and DCT method at floating point.

Dynamic Range, IQ and Tonal Depth Explained ( Olympus E450 owners)

Jargon Busting!


DR: Dynamic range. Seen a histogram? This is worth delving into on your camera display and at least on Olympus viewer on your PC. It relates very firmly to the actual digital data and the quality of the capture of the scene.

IQ No, your camera is not getting more intelligent, this is just a good old subjective reference "Image Quality"

Tonal Depth: this is just as important as dynamic range: given a hue or short range of light intensity, this is how many tones can be captured (and displayed!) such that the nuances of colour and shadow over shapes, textures and colour gradients like a sunse, add to the perception of quality and the technical accuracy of the shot..


How to Define Image Quality?

Well beauty is in the eye of the beholder and you could say therefore there is a dichotomy between technical accuracy and æsthetic value:

Most shots on the Dp forum showing Olympus DR etc are very dull subjects: standard test cards, studio objects or just desk foibles and views up the street from the office or shop with the new E5 on test!

However given some form of interesting content then you can say that the key contributors to image quality are:

1) representation of the scene in either true-to the eye or, a one that gives some æsthetic which impresses at first glance.

2) sharpness : usually the subject of the photo should be sharp: and when enlarged detail remains sharp, although this is a technical sharp: see the debate on how many mpx . I like many slightly blurry portraits - it is one of my own styles. Shaprness acheived by appropriate shutter speed, tripod, image stabilisation, correct focus and associated aperture mediated DOF:

3) Appropriate Depth of Field: the subject is either the only area which is represented sharp or the whole image is sharp ( by in large these two are what is regarded as contributing to good IQ.) DOF is a whole blogg on its own and can be mimicked in image software like gimp.

4) Accurate capture of shadows and highlights. The range apparent to the eye or a range which represents some æesthetic in the image.

5) A good representation of the colours and tonal depth of those colours across their hues and light intensities in the range between the white highlights and black shadows.

Rocket Surgeons Versus Picasso De Vanci

I guess the arguement could rage between the "technicians" and the "artists" on what IQ really is, but as in a previous blog on what makes a good picture, there are both technical elements and æsthetics which you could get rated by 1000 people and come up with an IQ rating if you must.

Certainly a good technical image will lend itself to representing the scene as intended or being worked up in after effects to a new æsthetic which may have been limited if the technical qualities were not there.

My favourite portrait photographers, including Anton Cobijn - the Joy Division and U2 image man, have often rather poor technical prints: very short DR, grainyness, blur, innapporpriate DOF , underexposure. However to me they are really striking.

Way Back When....

In the old days of film, given you had a reasonably good camera, there were two main areas affectting the quality of your colour prints.

1)IThe film as bought: how good the film was relative to its' ISO. it's colour "hue" representation, tonal depth and most of all the grain and noise.

2) At the Lab: The quality of the developing of the film ; the accuracey of the test exposure meter/print; the quality of the paper and chemicals, and the freshness of both.


Now we can't blame it on the lab or the shop or the film maker!

We have responsibility for the quality of our images, but it is luckily much easier : especially with Olympus which has such a good JPEG rendering engine and actually RAW files which are sensibly worked up from the totally raw data. Unlike some camera makes, you start with a reasonably sharpened ORF with a little more sharpening on the JPEG.

The 400 series do not have the absolute best DR and Tonal depth but for the price the overall image quality even with the so called "kit" lenses, is really stunning.


What is dynamic range and how do I see it in an image?

Well you see it most immediately in the histogram as mentioned, but wait up: take a look at the image and the exposure just after you have taken it, and compare it to what the eye sees. Try and alter exposure and focus until you capture near to what the eye sees.

Now look into the shadows and highlights of the image: there is a function in "INFO" which displays highlights and shadows ( I believe it is above 251 and below 5 respectively, more on numbers later.)

The key to understanding DR is that it describes these two extremes: what the camera can pick up in nuances before shadows become black and highlights become white.

Another related issue which is unfortunetly prominent in some conditions and exposures on the E450, is that in being devoid of nuances at the highlight or shadow end, flecks of noise, usually reddish, creep in. More on noise later.


Numbers:

Dynamic range is defined in digital terms as the range from 0 to 255, with 0 being pure black on the left of the histogram or curve (graph) and white highlight being 255 on the right. In black and white, everything in between is a shade of gray in a BW image for example.

Some cameras do not acheive a complete capture of this possible range: they will therefore lose some tones into being just black or white : so for example 250 upwards or 10 downwards. Also some exposures in the E450 will not capture the range: this would be under or over exposed for the situation.

Finally relating to this last point, some subjects or scenes either lack light intensities from the whole range: high key/ low key , or sometimes the sun is just too strong for the camera to "find" grades in tones from about 240 upwards. The same is true in very dark images, where there is not enough light / long enough exposure to give any detail in the shadows.

The relative proportions of image area which are captured in the image are then shown as the peaks on the y axis.

To capture the very best dynamic range for the conditions on the E450, I recommend using bracketed shots on programm or shutter priority, such that the aperture setting is changed. Further to this, if you want to capture the best exposure and "local" dynamic range for a given subject within the fram, like a face in a crowd, or a bird in a shaft of sunlight, then use the spot meter function. The camera will then interpret only the correct exposure and DR for that small area of the shot. ( see earlier blog)

Tonal Depth

This is actually not justt the y axis of the histogram : it is more the accurate representation of the spectrum of colours and their light intensity or saturation.

Shadows are black? Well many argue they are blue tinged, brown or infact grey and not pure black. But at first glance , total shadow point of below 5 on the scale will appear black to the eye.

The whole thing is easier to understand in the LAB system of colour separations than in the monitor oriented RGB scale. In RGB all the colours combine : all off, "0" black, all on 100% light intensity, 255 white. In LAB the blue-yellow and red-green primary range are separated and represented as the opposiste they actually are: where there is yellow, there can be no blue!

The third channel is pure BW, grey scale. In LAB you get a better idea of shadows and highlight areas, and how this helps add depth to an image, while you also have a better level of control in colour balance ( a whole other topic!)

It can be worth converting to the LAB system to work up some images: especially high key and low key, or subtle tonal range images and then finally converting back to RGB JPEG at the end. Nearly all monitors are RGB so you see what you are going to acheive when the "separations" are viewed together.

Why is Tonal Range Important to Me?

What you can do is refer to books which discuss and illustrate "polarisation". This shows how good tonal range is necessary to help with depth and sharpness in a pciture and perception of good image quality. You can reproduce it by super compressing a JPEG at a small size at 50 ppi. Save a copy of a portait ( face only) and then repopen it to see the poor quality and then look at the historgram.

There are large areas of the histogram with zero values: colours become clustered into individual peaks and there is an appearance of "super pixelisation" , box like areas of single colour. There is a sharp grading between colour or shadow areas, with contours often appearing on faces like some maps and weather charts!


Think about making this gradually better, in more detail , and then you start to realise that tonal depth is important for representing a 3D scene. Also it is important for displaying hues and saturations of colours across the historgram. The graph would have more peaks and a rounder curve between peaks given sublte grades of colours, such as in skin tones or fabrics with light coming from a low sun for example.