More on post processing

[RVT1K]

I have a question concerning the "need" for post processing but it is based on what may be bad information.

That information is that your eyes have better dynamic range than your camera. Meaning your eyes will see more detail in a mixed exposure, especially in shadows.

And, based on that, post processing would be "necessary" to produce an image equivalent to what you actually saw.

Truth? Obsolete truth? Utter BS?

 

[Tim Tucker 2]

Utter BS...



People think in absolutes and so assume that to look the same you have to keep it the same. But it doesn't work like that because the DR of your output screen is generally much lower than the eye and so the eye will not see it as a high DR scene, it will see it on a device with low DR. This causes problems with your perception of colour and contrast. If you photograph a high DR scene and then try to display *absolute* detail in a still image in the far lower DR of your screen you will find that your impression of both colour and contrast will be greatly diminished.

Then you have to remember that the eye scans and adjusts as it scans, (it is not a still with absolute DR measurments), so it will construct an image which will have a greater *overall DR* but one that still retains the higher *local contrasts*. If your eye looks at the brighter parts it can't see the shadows and sees a high contrast between the two, when you look to the shadows you lose detail in the highlights but still see a high contrast. The resulting *composite* your brain puts together includes the impression of both the high DR and the high local contrast. On a computer screen your camera does neither, it reduces the overall contrast between the brightest and darkest to the lower DR of the screen and compresses local contrast so it fits within this space. You can keep the detail in the shadow but you lose the impression of contrast, colour, depth and with it light, or you can keep the impression of contrast, colour and depth and lose some of the detail in the shadows.

This is the problem, you have to change the DR, or the contrast/absolute difference in brightness, to display it on a computer screen or in print. Changing this makes the image look different. You can up the saturation which will increase the contrast between the colours, and you can block the shadows a little and it will look similar, but it won't be the same.

There is no narrative of *absolutes* where the camera captures the absolute information you see that because the eye simply doesn't work that way, it sees the scene differently and so a photo on a computer screen will always be different. In controlled lighting where the DR of the actual scene is kept to 9 or 10 stops then it transfers well to screen or print, if you try to capture a larger DR you run into the problems above.

 

[dxqcanada1]

Hmm, the human brain interprets information fed to it by the ocular sensors and creates a "picture" from all the data that is continuously being analyzed which includes other input that will influence the "interpretation" ... this causes much anguish when we humans try to present that "picture" into a form that can be viewed by others.

 

[wfooshee3 2]

You can see a wider range of levels with the eye than with a camera, but that's not because the eye has a wider range, it's because the eye adjusts to what you're looking at. As you look at the dark corner of the room, the light from the window is not on the center of your eye, and your pupils can open up. Nothing near or outside the window is even perceived, much less visible. Then you shift your gaze to the window, and the cat on the table under it. What was in that dark corner is no longer perceived in the brain, as your eye has adjusted for the light on the scene you're looking at now.

If you try to take a picture that includes the window and the dark corner, there will be a compromise. A single exposure just won't cut it.

So you can see more detail in a scene with wide lighting variations, but not all at once. You see what you're looking at, and your eye adjusts for the center of its image. You can't even make out the title of a book that's in your peripheral vision, but look at it, and it's plain. You can see the cat, or the dark corner, but not both at once, even if both are pretty much right in front of you.

That action is much different from photographing a scene. The camera can't produce an image that adjusts itself to the point that you're looking at; it only gathers the light as directed by the photographer or the light meter, whichever is controlling the exposure. The camera does have a limit to its dynamic range. there is a limit to how dark and how bright items can be in a single exposure.

You can use HDR techniques to combine images that are bracketed around a median exposure and combine them in software to produce an image that a single shot just can't get. The software uses the images to gather detail from the light areas and the dark areas, keeping all visible in the final image. HDR stands for High Dynamic Range, but the result is actually reduced dynamic range; you're compressing the dynamic range into what a single image can contain.

An example, if you'll indulge me. My girlfriend's cat in a garden window. The first three images are shots exposed two stops under the median, at the median, and two stops over. That way, I can capture the trees and stuff outside in the underexposed frame, and capture the fur texture in the darkness of the cat in the overexposed image, and the software can produce the 4th image as a result. To me, THIS is what HDR is all about; recovering excessive light/dark differences into a single image, and not the metallic, painted look that so many people use HDR to get.

Median (matrix-metered) exposure:

2 stops underexposed:

2 stops overexposed:

And the final processed image:

The final image represents what I felt I was looking at as I sat in the room, but it's not possible to capture that image with a single exposure; the range of light in the scene is simply to wide,

 

[Strodav]

You might find this article from Cambridge Colour useful: Cameras vs. The Human Eye

Local adaptation makes your question difficult to answer. Say you are looking at an hdr (high dynamic range) scene. When you look into the shadows with the center of your eye, the iris will open up a bit allowing you to see into the shadows. When you move your eyes to look at an area with highlights with the center of your eye, the iris will close down a bit. So, what did you really see? The way we handle those situations in photography is to chose to blow out the highlights or shadows, or use hdr techniques, i.e., bracket your shots then combine them effectively compressing the dynamic range of the original scene to fit your system's dynamic range - the number of bits you use.

 

[weepete]

I'm not sure you guys assumtions are correct. I can definately see more dynamic range than any camera I've used, even when I don't move the focal point of my eye I can still see details in highlights and shadows that the camera won't record. Particularly in backlit situations.

I do think that there's a thing where digital cameras record more information in the highlights, where as human vision is more geared towards seeing information in darks, and our brain activley tries to make asscociations between shapes and forms we know (that's why we can see complex shapes in clouds for example).

 

[Vtec44]

Post processing can be a variety of things and it’s all personal to the editor. So some people may want to duplicate what the human eyes can see while others want to crush the shadows and blown out the sky. There isn’t a single way to post process.

 

[RVT1K]

You might find this article from Cambridge Colour useful: Cameras vs. The Human Eye

Local adaptation makes your question difficult to answer. Say you are looking at an hdr (high dynamic range) scene. When you look into the shadows with the center of your eye, the iris will open up a bit allowing you to see into the shadows. When you move your eyes to look at an area with highlights with the center of your eye, the iris will close down a bit. So, what did you really see? The way we handle those situations in photography is to chose to blow out the highlights or shadows, or use hdr techniques, i.e., bracket your shots then combine them effectively compressing the dynamic range of the original scene to fit your system's dynamic range - the number of bits you use.

The article was a good one.
Thanks for the tip.

 

[Tim Tucker 2]

I'm not sure you guys assumtions are correct. I can definately see more dynamic range than any camera I've used, even when I don't move the focal point of my eye I can still see details in highlights and shadows that the camera won't record. Particularly in backlit situations.

Excuse the essay, but it can be a toughie to get your head around.

You appear to be making an assumption yourself, that what you see is what is projected on to the back of your eyes, the raw data. But what you see is processed data and appears to be a direct feed in the same way the image you look at on the screen appears to be from the position of the camera. For instance the data you see is two images from slightly different positions processed and overlaid into one single viewpoint and understanding over a period of time... It is information that comes from the eye which is processed in the brain where a composite understanding is produced. In much the same way that you can't see the raw data as the camera sensor sees it but instead see the result of it processed into a jpeg and displayed on the screen on the back of the camera.

There is a rather neat illusion that demonstrates this, where you concentrate on a fixed point in the middle of one image then switch to the point in the middle of the other. And it shows quite clearly that your eye has learnt and adjusted what you saw in the first image and is applying that correction to the second, at least initially.

But you seem to be trying to analyse what you see as if it were a fixed capture with absolute values that you can see accurately.

We are not born with the knowledge that allows us to recognise trucks and cars and be able to estimate their distance, we learn this through experience. It is the same with learning to recognise the conditions and light, we are not born with it but learn through experience. How do you tell a sunny day from a dull and overcast one? Actually it's mostly to do with recognising the patterns of local contrasts. In the real world we can learn and tell them apart with ease because they are normally quite consistent, it's only in images where they change. For instance, if you tone map a sunny scene image by increasing the shadows and decreasing the highlights then you are in reality reducing the differences in local contrast between the sunlight and the shadow, making them more uniform. Then when we look at the image we recognise a pattern that is more consistent with an overcast day than it is a sunny one and interpret the information as though it were an overcast light.

This is the main problem, the data is not absolute in an image it is just a pattern of coloured dots on a screen that you interpret against similar conditions you've learnt to recognise through experience. There is no objective reality in an image, it is all representative, it just *looks like* and is never *the same set of values as you captured*. In the real world the actual difference between the lightest and darkest parts of what you see varies from a sunny day, to an overcast one, to being indoors under artificial light. In an image on a computer screen it is fixed, you are using the same range of brightness for all three images. If you look towards the sun then you will either shield your eyes with your hand or screw your eyes right up, you do neither when you look at a photograph taken into the sun. The fundamental brightness and differences in brightness of the scene have all changed in the image, they must change to fit the brightness range the screen outputs. It simply can't display the information in any way close to the absolute levels in the real world.

You can see this with the *colour on an overcast day* effect, it isn't more saturated at all. Colour looks far more saturated when it is presented in bright sunlight against the bigger and more abrupt contrasts, not the soft contrasts of a dull day. But many photographers think in absolutes, because the colours look more vibrant they actually are more vibrant. Here's what really happens...

On a bright day you have a very high DR which produces a high *absolute* local contrast. The bright pale yellow flower reflects a lot more light than blue one in the shadow behind it. You understand the shadow detail to be dark but can still see it and your understanding is it's blue and the yellow is that it's much brighter. The local contrast is high and it affects your perception of how bright the colours are. Now convert that image to a computer screen that has a fixed and much lower brightness range. If we try to maintain the impression of brightness by keeping as close to the actual contrast or difference in brightness as possible the pale yellow is pushed more towards white and the blue more towards black. If we try to maintain the absolute colour then we have to reduce the brightness of the yellow further and increase the brighness of the blue and so now the absolute contrast between the two is so reduced that the yellow no longer appears so vibrant and bright against the blue and the blue appears more pale against the yellow. So we increase the saturation to regain some contrast between the colours.

Now this range of brightness also applies to the point of capture, it's not just the screen. Your bright pale yellow being near the limit of the cameras capability will record as closer to white and the blue will record closer to black. What happens on an overcast day is that the difference in brightness between the yellow and blue reduces, so now when you expose (exposure is the act of reducing the variety of different brightness of different lighting conditions to the same fixed range of values that your sensor records, it reduces all scenes to the same range of brightness), the colours move away from the black and white more into the center of the histogram and so into the range where colours record well. And because you don't have the same difference in shadow/highlight across the a single flower the whole yellow flower records colour that is more *yellow*, (in harsh sunlight a yellow flower will record values from near white to a dark yellow, in diffuse light all values will be nearer a bright, vibrant yellow. You have recorded more colour and less of it is close to black or white simply because when you set exposure more of the colour falls within the range of the sensor to record, but the image looks flatter because it lacks the contrasts that give the appearance of volume. This is another problem in images with fixed values, to display depth you use shading which tints colours towards black or white, reduces colour. To have colour you must decrease this shading and so the image appears flatter).

In none of these examples is the image on the screen the same as what you saw, it's impossible to reproduce exactly what you saw when you use a different range of brightness, colour and contrast to display it. Prints and screens have a fixed range of brightness that they use to display all images and there is an optimum scene DR that transfers well into this space. If you exceed that then you MUST by definition change the relationships between brightness, contrast and colour to make it fit into that space, it's impossible to dispaly a high DR scene without doing this. And when you do this you also change the pattern of local contrast and so change the way we see and interpret the data. The more DR you try to cram in then the more abstracted the image on you screen looks, the more you change the values and the relationships of brighness/colour/contrast then the more abstracted the image looks. There are ways around this that involve perserving local contrasts over global ones, or changing values in a non-linear way. But most of your digital tools are linear and work on global values and this is why many high DR scenes look so abstracted when processed and viewed on computer screens or in print.

There, simple...