Keeping Detail in the Dark Depths

This is a very dark shot, in a very dark room, with no flash (intentionally)

I think I managed to keep all the detail in the darkest parts of the image, for me it was a technical exercise (this is the one image taken) so don't worry about the composition, although I love the look of the microphone in this image.

Click for a 100% version (saves me posting crops, and you get to scroll around all you want.)

So the feedback I'm looking for is, have I nailed keeping details in the shadows finally. That is all this image is about, nothing else.

IMAGE LINK: http://www.flickr.com …denby/284108873​0/sizes/o/

I think you are kidding yourself about the shadow detail.

Use the histogram, not your eye, to determine where the detail is and where it is not. That's what it's there for.

It shows about 17000 pixels at value 0 and about 60000 pixels at value 1. Those numbers represent about 3% and 10%, respectively, of the total count of pixels in the image. I would describe that as having the deep shadows blocked up.

So, I'm curious, to shoot for that, would you expose 'normally' then drop exposure back in post?

(Like shooting day for night @ dusk......)

Sorry but it just looks underexposed (and noisy) to me.

The details that I see in the microphone are being produced by a few mid tones and highlights. There is a lot of real estate that it completely black and has no details that I can see. I think that what you are really after is a dark image with sufficient mid tones and highlights to reveal the details. In that respect, I think that you have done a good job of achieving that goal.

Thank you for the feedback so-far.

Robert, 2 points.
1) Normal viewers use their eyes not histograms
2) In the image I've added a black border, and a white border. These will confuse a histogram somewhat, so I fully expect those pixels to show up in your histogram right on the 0 and 1 values. The fact the microphone stand is pure black and has no real life texture to it, will probably put a lot of pixels in that region also.

If you take that into account and you still think their is an issue please let me know.

Walter, it is a little noisy, i was surprised as it was shot at iso400, its definitely noticeable on the 100% version, but I think on a print or a smaller resize it would be ok. I have not applied any noise reduction so-far as I'm not intending this as a final image, and I'm between noise reduction tools at the moment.

Do you think its iso noise or because of the dark exposure?

Bill, that's an interesting observation, you might be on to something there.

I pulled your image into my computer and your web image does not do it justice.
The histogram I show was from an after crop image, got rid of the frame.
.

HTTP response: 404 | MIME changed to 'image/gif' | Byte size: ZERO | PHOTOBUCKET ERROR IMAGE

KarlosDaJackal wrote in post #6270344
Walter, it is a little noisy, i was surprised as it was shot at iso400, its definitely noticeable on the 100% version, but I think on a print or a smaller resize it would be ok. I have not applied any noise reduction so-far as I'm not intending this as a final image, and I'm between noise reduction tools at the moment.

Do you think its iso noise or because of the dark exposure?

I'm no expert on the subject but I believe you'll generally see more noise at a given ISO in underexposed images.

Thanks Chauncey, that's levels displays as a histogram which usually looks a bit different and does not give a count of pixels, thanks for using your eyes though, I'd thrust them much more.

I don't fully understand Roberts numbers. The original full size image from the click link, with borders has 11575926 pixels. If I use Roberts number (17000 and 60000) he is telling me that I have 3% and 10% gone, but the maths says 0.1% and 0.5% that's a huge discrepancy and I hope the method used is explained so we can learn something from it.

My maths using Roberts numbers
17000 / 11575926 = 0.0014 * 100 = 0.1% (too white according to Rob)
60000 / 11575926 = 0.0051 * 100 = 0.5% (too black according to Rob)

My histograms
My histogram with the black inner border of 8*12 and white border of 80*120 still present say, says that 1.5% of the colours of the image are value 0, and that 5.4% are at value 255. That seems like a good compromise to me in this sort of image, if you can't see that detail with your real eyes in that light, how will the camera see it. I'll attach my actual histogram reading.

But honestly if the thinking in this part of the forum is that we should all come up with the perfect histogram for each and every shot, and treat each image as 100% science then we all may aswell be shooting on green box mode.

My feeling is cameras do not have the same dynamic range as our eyes, and HDRs are not practical in a lot of situations, so you need to look at the image and judge what you see. Just the same as DOF can be used to draw attention to what the human eye would have focused on in a scene. If someone came up with a sharpness resolution chart for each image and told anyone with a narrow DOF, that only 5% of your image is sharp so bin it, we would all laugh at the suggestion.

Darkness and Light, Details in Lowlights and Highlights should be just as much a tool in photography as DOF in my opinion. I have a book with 500 of the most famous photographs, and I'd say a good 50% of them would not pass the histogram test of this forum.

Time for a unrelated "bright" image distraction from all that science and math stuff.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

Sorry about the confusing numbers, but I attribute that to Adobe and take no credit for it. The histogram display in PSCS has some mysteries that Adobe doesn't seem to want to explain. Their histogram uses a smaller image than what you give it, and I have never been able to figure out why it is the way it is. Another confusion factor is that the number of pixels in the "Count" is for the total of all three channels whereas the value of "Pixels" is only one third of the total number of pixels that are accounted for in the "Count". Be that as it may, if you use its numbers judiciously, it is a valid analysis tool which is better than the eye alone, which is somewhat subjective. Notice that I do not say better than the eye. I am saying better than the "eye alone".

I avoided the borders when I took my data from your image by doing a selection of everything just inside the frame.

P.S. I used only the smaller picture posted here rather than the linked image. The linked image has radically different values in the first few bins.

Robert_Lay wrote in post #6272056
Sorry about the confusing numbers, but I attribute that to Adobe and take no credit for it. The histogram display in PSCS has some mysteries that Adobe doesn't seem to want to explain. Their histogram uses a smaller image than what you give it, and I have never been able to figure out why it is the way it is. Another confusion factor is that the number of pixels in the "Count" is for the total of all three channels whereas the value of "Pixels" is only one third of the total number of pixels that are accounted for in the "Count". Be that as it may, if you use its numbers judiciously, it is a valid analysis tool which is better than the eye alone, which is somewhat subjective. Notice that I do not say better than the eye. I am saying better than the "eye alone".

I avoided the borders when I took my data from your image by doing a selection of everything just inside the frame.

I am not familiar with the program that you use above for histogram. What is it?

Its the GIMP, which is what I use to do my PP. Unlike PS, if you want to know how the histogram works, its either documented, or you can email the guy who wrote that part, or you can read the C code yourself.

FYI, i use a thing called UFRaw to get the image into GIMP and it has histograms of its own that show you in percentages the number of over and underexposed pixels live. You can even hold down the underexposed button and it will turn the whole image black, with any underexposed pixels highlighted in whatever primary colour channel they are clipping in (usually green first), same for overexposed only the base colour is all white, with primary colours showing the hot pixels.

I can't use it on my work laptop but I could get you a screenshot later, if you like.

KarlosDaJackal wrote in post #6272091
Its the GIMP, which is what I use to do my PP. Unlike PS, if you want to know how the histogram works, its either documented, or you can email the guy who wrote that part, or you can read the C code yourself.

FYI, i use a thing called UFRaw to get the image into GIMP and it has histograms of its own that show you in percentages the number of over and underexposed pixels live. You can even hold down the underexposed button and it will turn the whole image black, with any underexposed pixels highlighted in whatever primary colour channel they are clipping in (usually green first), same for overexposed only the base colour is all white, with primary colours showing the hot pixels.

I can't use it on my work laptop but I could get you a screenshot later, if you like.

Many thanks for the additional information.

Attached are an image of my raw converter, and an image of my raw converter when the underexposed button is held down.

It shows exactly each pixel that is out of range. I don't think those pixels are significant to the whole image.

HOSTED PHOTO
please log in to view hosted photos in full size.

HOSTED PHOTO
please log in to view hosted photos in full size.

The screen shot which shows the "underexposed" pixels looks amazingly like the shadow clipping feature in Adobe Camera RAW. If I make the reasonable assumption that those pixels shown in that screen shot are pixels that were clipped to value zero, then it follows that the information in those pixels has been lost.

In your original post you said that you thought that you had kept "all the detail in the darkest parts of the image."

Robert_Lay wrote in post #6273902
The screen shot which shows the "underexposed" pixels looks amazingly like the shadow clipping feature in Adobe Camera RAW. If I make the reasonable assumption that those pixels shown in that screen shot are pixels that were clipped to value zero, then it follows that the information in those pixels has been lost.

In your original post you said that you thought that you had kept "all the detail in the darkest parts of the image."

Those pixels (all 0.4% of them) are value 0, not necessarily clipped to 0, also they are all on a pure black microphone stand so they should be 0 or very close. Also, I don't think any of them few pixels are "details". If you notice the gaps between those dots, are not clipped and outnumber the other dots, so you still see detail.

I think this thread/discussion has proven if nothing else that the tools you use to evaluate other peoples pictures are less reliable than people eyes.

If you go take a picture of a something that is painted black, in low light, with shadows falling on it, and end up with an image with 0.0% pixels of value zero, I'd say something was very, very wrong.