[PlayRaw] Chrysanthemum flowers in BW

I hope there is some interest in this. I didn’t want to post on the original post because it was so old, so I opened this. Referring to @XavAL, Xavier Bartol’s post, Chrysanthemum flowers, I played with BW development, and also was trying to learn what the difference is in how you bend the contrast equalizer curve, and I have a couple of results.


_DSC0488_13.NEF.xmp (49.6 KB) dt3.3
and


_DSC0488_14.NEF.xmp (60.4 KB)

The difference between them is with 13, I curved upward the coarse and fine ends of the chroma and edges curves in contrast equalizer, and in 14, I curved up the middle.

I am interested in how others would tackle the RAW, and if anyone has experience on printing BW, what observations they would make on how well these would print, or if something more needs to be done to make good BW prints. My feeling, from looking at the jpgs on my monitor, is that 14 would print better than 13.

4 Likes

For convenience, the raw file is here.

1 Like

I would do whatever accentuates the subject.

Especially noticeable viewed large, both images have weird black fringing around the top of the white flowers. What caused that?

Hazarding a guess, it might be a masking issue.

image

It’s the chromatic aberration correction changed to monochrome withe the contrast turned up. Turned on…

Turned off…

2 Likes

Get a couple of 6x4 inch prints done at the local chemist, it’ll cost about a dollar. “better” is subjective, “reality” is probably xmp number 13, where the yellows translate to a mid-tone, number 14 with the added 20.53% contrast is a much whiter flower.

Probably some things to consider, computer screens are light sources, prints are reflective, so a print is very much dependent on how it’s viewed. Monochrome is not reality so pumping the contrast for subject emphasis is much more acceptable, but then one has “opinion”. Generally people say that a perfect print has a range of tones and detail visible, but it easily becomes subjective, how much detail in the shadows? is it okay to have to look for it in good light?

I have a story, I started in the darkroom for a group of regional newspapers. Early on, I did a reprint for a member of the public, a 10x8 of a black rabbit next to a white rabbit, it was possibly a little too contrasty but you could certainly look at it and find the detail in the white and black fur of the respective bunnies.
A week later the town office sent it back, requesting a reprint because the customer was not happy. This I could not get my head around, I printed it again, slightly softer where the whites were what I considered to be light grey, the black rabbit was losing it’s depth. A week later the print came back, this time with a copy of the photo in the newspaper and how the client wanted the print. Newspaper paper is grey, there’s no “white”, there’s no depth to the blacks. I actually got the lady into the offices and dug out the original 6x4 print that had gone into the paper and tried to show her “black rabbit, white rabbit”, newspaper print = dark grey, light grey rabbits. So the “correct” print was the original and my subsequent enlargements. The lady was not having any of it, she wanted a print exactly the same as what had appeared in the paper, even though she physically owned the two rabbits, one black, one white. I printed a very soft version eventually, there is no moral, just a commentary that it’s all personal preference.

White is also comparative, put a white envelope on a plate with a paper towel and the differences in whiteyness are visible. Which is why I suggest getting some prints knocked up you may find that the preferred screen version has blown highlights and blacks too deep in hand. Not blowing the highlight detail, you can see a lot more dark detail holding up the print and squinting, but you won’t magic detail in white areas if they exceed the printers capability.

2 Likes

Thanks for all the feedback, especially about the chromatic aberrations. It’s not that I want to print this particular picture, but that I want to learn about making good BW images. My daughter went on a trip out west. Not knowing about shooting in RAW, she shot just jpg. She sent me several she was interested in getting printed. I converted them to TIF and did some editing. One, I changed to BW, and she really liked it, but when I printed it, it just didn’t have the look that we wanted, and I didn’t know what to do to get it. I think @nodal is right about getting images printed locally. I won’t improve until I getting printed results and then modify them until I get something I like. That’s how I will learn.

I’ve been intermittently playing with monochrome, and the freedom to do stuff is rather relaxing compared to my colorimetric work of recent. Thing is, I think it’s important to start with a colorimetrically “right” rendition as the starting point for monochroming, as it is the mix of the channels to produce shades of gray that starts one with the rendition. Here’s my take on it in that regard:

I started with the linear RGB which is already quite nice color-wise, just a little flat in tone. Cropped to put the yellow bloom where i wanted it, then did the grayscale conversion like this:

_DSC0488-rawproc-gray

Using predominantly the green channel gave the most separation between the yellow bloom and its purple siblings. A bit of green helped, and also upped the overall tone.

Next, I added a RGB control-point curve to pull the darker tones further down:

_DSC0488-rawproc-curve

I just love to push those shadows into obliviion, adds mood to the image. I’ll do this for both color and monochrome compositions.

I’ll also play with re-introducing color in the form of toning the dark parts with a per-channel control point curve where the low-left control point is pulled right or up along the axis so very slightly. This image did really nicely with a subtracted red channel, giving dominance to the blue and green components and rendering a cooler tone; I didn’t include it here because presently I’m getting a color difference between what rawproc renders and my browser that I need to get control of…

Simplistic stuff, but It’s just so much fun… :sunny:

5 Likes

The principles in this 8 year old article are still relevant and it has multiple linked articles on other conversion methods.

Personally I think the correct balance would be between the two images, I suspect that the more contrasty one would clip the tips of the petal in a print though I prefer it on screen and naturally I may be entirely wrong :slight_smile:

@nodal The article was very helpful to me. Thanks for pointing that out.

@ggbutcher Did you adjust exposure to get a good middle gray? I took the RAW as is, except for the standard modules, set channel mixer RGB as shown and set up rgb curve:

image

But my result does not have nearly the tonal variation of yours:


_DSC0488_15.NEF.xmp (20.8 KB)


_DSC0488.NEF.xmp (10.1 KB)

2 Likes


_DSC0488.jpg.out.pp3 (15.9 KB)

@ggbutcher’s contribution is very nice; Much overall detail, but especially the white(yellow) flower and an almost palpable softness. Great edit!

3 Likes

Curve tools are not created equal, so to speak. Different spline algorithms affect the slope of curve segments, and that’s what you’re after, a steeper slope in the range occupied by the bright flower. That, and I think your histogram is depicting the display data, so it requires different treatment with the curve.

What you want is more steepness in the upper part of the curve, which is where the bright flower resides. That’ll give more contrast to those tones, and thus more tonal variation. Also, the lower control point probably needs to move right to crush more of the lower tones.

Thanks! Looking at it on the ‘big screen’ I wish I had brightened it just a bit more. That, and I figured out my color difference; using a bad display profile, danged color management… So, with ‘bit brighter’ and a base subtraction of a smidge of red from the lower tones to put a blue-green tint to them:

Here’s the per-channel red curve to do this:

_DSC0488-rawproc-tint

Note 1) red channel selected, 2) lower-left control point is scooched to the right just two units, can’t really see it behind the dot but the coordinate of the selected control point is at the upper left.

About printing, I haven’t really done much since moving to digital, but I think I’d print this on regular matte color paper.

2 Likes

Most of my edit was about adjusting the brightness of the flower and not the rest. Arguably one of the hardest thing to do in this one, it’s a bit harder then just upping the exposure slider…

That toning makes it rather special; Lifts the flower just that much more compared to a full B&W based monochrome image. Do a back-and-forth between them and see how much more 3 dimensional it looks al of a sudden. Like it!

Ah, that’s the beauty of this image, the subject is so different in both tone and color from its surroundings. @underexposed’s histogram tells the story; the purple background is the lower hump, and the yellow subject is the upper hump. All you need is to put some steepness in the segment of the control curve that spans the interval between them, and that’ll guarantee some amount of tonal separation between the two…

Yes, exposure isn’t going to spread things out, as you’re just multiplying the entire image by a single number. That’s why exposure isn’t considered a non-linear operation; magnitudes change, but the relationships are maintained.

There are just some things you can’t easily do with filmic… :stuck_out_tongue:

Again, thanks for the guidance and the edits. I followed what you said, Glenn, and managed to acceptably match what you did, which, I agree with Jacques, is very well done.

My curves ended up a long way from looking like yours. My x axis on the red channel ended up at 25.8.

Maybe this is the wrong place to ask. Be that as it may. I use IrfanView to display images on a Windows 10 machine. Looking at all these B&W images in dt first and then as jpgs, it appears that the black point of the images is lower in IrfanView than dt. Since I calibrate, I wonder where that comes from: exporting from dt, IrfanView, or something else? Any ideas?

Knowing the ways of the curve is good. Knowing that every point on the spline is a lookup reference for transforming an X into a Y helps to know what’s going on at that point, and maps into what is happening to adjacent values. Shallower slope=less contrast, steeper slope=more contrast, etc. This applies to any tone transfer function, especially filmic. I was glad to see @aurelienpierre include a depiction of the filmic curve in darktable; I understand what that tells me about what it’s doing to the image…