A book report — proofreading

This is a continuation of a series of posts that I started what seems like a long time ago about getting a book designed and published. The series starts here.

I’ve already posted about how I’m proofing the images for the book. However, there’s a less-fascinating (to photographers) but still important, proofing task: all the text. There are five main conceptual groups of text in the book.

  • A forward, written by Brian Taylor
  • An introduction, written by moi
  • An afterword, also by me
  • The captions
  • All the miscellaneous tittles, copyright notices, credits, etc, sprinkled throughout the book, the dust cover, and the slipcase.

I proofed all the text for the first three before I sent the copy to Jerry as three Word documents. Brian’s writing style is quite different from mine, and I didn’t do any style edits on his piece. I dealt with the captions by putting them into the metadata of their respective images before I sent the images to Jerry. I figured that would keep each image associated with its caption, eliminating the embarrassing and frustratingly difficult to find error of captioning the wrong picture.

Jerry and his people put the text and photos into an InDesign document, exported it as a PDF, and sent it to me for proofing. I proof better from hard copy than from a computer screen. I don’t know why, but I just see little things when I’m looking at paper, things that might easily escape my attention on the screen. So I printed out the PDF. I immediately ran into a problem. The pages in the PDF were full spreads — what will become two facing pages in the printed book. Since there are many images that span two pages, this is a good way to visualize the aesthetics of the book. However, when a double page is printed on the largest non-photographic printer I have (8.5×11 inches) almost all the text is illegible, at least to these tired old eyes.

Jerry reformatted the PDF into pages, and things were better, but still not great. I had special difficulty with the captions, which are printed using a 50% gray spot color, so they don’t distract the viewer from the image itself.

I had to proof the captions on the screen. When I did, I found some of them were with the wrong image. Apparently, rather than having InDesign place the captions automatically from the metadata, Jerry and his designers placed them manually. I also found errors in my own captioning in Lightroom.

When all was said and done, neither Jerry nor I had high confidence that we’d gotten all the mistakes worth finding. Jerry offered to find a proofreader. I said yes.

Five hundred dollars and two weeks later, Jerry had a hard copy printout of the book marked up by the proofreader. She found a lot of errors, some of which I wouldn’t have bothered fixing, and many that would have made me blush had they found their way into the final book.

I wanted to see what the proofreader had done, but at the time I was housebound. Jerry’s designer had made the corrections in an InDesign file, and apparently InDesign doesn’t have an equivalent of the Word revisions tools. In any event, Jerry couldn’t email me the markup; if I wanted to see it, snail mail was the only choice. I decided not to do that, and just did an on-screen proof of what the proofreader had done. She made many changes that were hard for me to find, which I guess means that she did a good job.

 

 

 

A book report — coloring outside the lines

This is a continuation of a series of posts that I started what seems like a long time ago about getting a book designed and published. The series starts here.

A lot has happened since I last reported on this project in — gosh, has it been that long? — July. In this post, I told you about deciding to have Jerry Takigawa do the design, and Brooks Jensen handle the project from there, with Hemlock doing the printing.

Jerry did a marvelous design, with striking double black pages separating the sections devoted to each city. He also designed a spiffy semi-transparent fly sheet. I sent a pdf to Brooks, who got back to me with some questions. Most were easily dealt with. A few were not.

Brooks pointed out that the all black facing pages would rub against each other as the book was read, acting sort of like sandpaper, and causing the nice rich black to deteriorate over time, to say nothing of causing black dust to get all over everything near the book. Jerry said that he planned to deal with that with an aqueous varnish. But the press that Brooks was planning on using didn’t support that. To get the varnish, we’d have to go to a different press, which upped the cost a lot.

In addition, Brook’s standard book had nothing like Jerry’s attention-grabbing flysheet. If would have to be glued in by hand, at not-inconsiderable cost.

Finally, after getting a quote from Hemlock for doing the book Jerry’s way, Brooks said that this was turning into a custom book, and that’s not what he was selling, and suggested that I have Jerry work directly with Hemlock.

I had to agree. I really like what Jerry has done with the book, although at the time I didn’t realize that he was designing something that didn’t fit Brooks’ specs. I’m probably not going to do this a lot, and I’ll just hold my breath and write a bigger check.

Sony a7II CDAF anisotropy?

A poster on DPR claimed that the contrast-detection autofocus (CDAF) and the focus peaking on the Sony a7x cameras are the result of the same hardware and software. I begged to differ, and offered the focus-peaking anisotropy that I’d demonstrated earlier as evidence. He said that the CDAF did that, too.

Hmm… Looks like something I should test.

I took a Sony a7II and turned off phase-detection autofocus (PDAF) which would confuse things. I put the Zony 55 on it, and set it to wide open. I set the ISO to 100, pasted a target that looks like this to the wall:

MATLAB Handle Graphics

This target has sinusoidal modulation with an amount of contrast that varies in a uniform fashion, and also has the modulation frequency changing in a predictable way.

I put the camera on a tripod, set the focusing to Flexible Spot, AF-S, turned the light down to get the exposure to a few tenths of a second, and looked for the places on the vertical and the horizontal target where the camera could just barely focus. There was some difference between the two target orientations, favoring vertical lines.

However, the CDAF was not at all like peaking, where horizontal lines produced no peaking at all. The CDAF was very capable of focusing on parts of the target containing only a single low contrast horizontal structure.

Therefore, I conclude that the mechanism by which the peaking indications are generated is not the same at that which performs CDAF.

 

 

How to get started in underwater photography

I’ve been out of the underwater photography game for a decade now, so it doesn’t happen very often anymore. But people who knew about my hobby used to come up to me at parties or corner me at some tropical resort and ask the question. Or sometimes people on dive boats, inexplicably undeterred by watching my toting, fixing, and cleaning of heavy, expensive, delicate, and fickle equipment, would ask it:

“How do you get started in underwater photography.”

Schooling Barracudas

I had a stock answer. Nobody ever liked it.

“First, get a hundred dives in your logbook.”

Sometimes, they stuck around to hear my reasons.

CLOWN2

First, there are the dangers to yourself and your buddy.

Perhaps the biggest dangers are to your buddy. Photographers as a group make terrible buddies. Being a good buddy means having good situational awareness; knowing where your buddy is and if she needs anything at all times. That’s hard enough if you aren’t too experienced and have no camera in your hand. It becomes nearly impossible if you’re trying to make pictures with little experience. In fact, photography demands so much attention that I don’t think you should be taking pictures underwater unless your buddy has 100 dives, too.

Severns and trumpetfish

Even setting your buddy’s well-being aside, there’s keeping yourself out of trouble. Keeping track of where you are and how to get back to the boat is sometimes tricky if you’re paying full attention. People who aren’t distracted by a performing moray have been known to lose track of their nitrogen loading, their air supply, the changes of tidal current, and all manner of details that are important to survival and not ticking off everyone on the boat with a trip to the chamber or a search for your sausage.

FEADUST1

Then there are the dangers to the beautiful places where you dive. Inexperienced divers lose track of their buoyancy and crash into things even without a camera. Buoyancy control has to be second nature before you encumber both your mind and your hands with photographic equipment. If you are trying to set up for a shot, you are much more likely to flounder around and stir up the bottom than someone who’s just sightseeing. When that whale shark shows up and you’re trying to get the camera set properly, unless you know what you’re doing the current is likely to slam you into the reef.

ORANG88

Divers, you probably get the idea. But what if you’re a snorkeler?

Same thing, only maybe more so, since snorkelers – not the ones who free-dive to 60 feet, and spend 2 minutes down there, but the mere mortals – often find themselves in conditions where there’s a lot of surge. If you’ve get a reg in your mouth, you can go down and get away from it. If you’re a snorkeler, you’re stuck with it. That means there’s even more of a chance that you’re going to get tossed about by the ocean, and, if you have a camera in your hand, even more of a chance that you’re going to tear up the reef.

Orange Clown Green Anemone

Does Lr export color space matter in Macbeth testing?

This is the 33rd in a series of posts on color reproduction. The series starts here.

There are a few loose ends in this Macbeth chart analysis work. One is: does the color space used to export the test image from the raw developer make a difference?  We already know that it does in the case of sRGB, for which one of the patches is out of gamut with many illuminants.  The chart is within gamut for both Adobe RGB and ProphotoRGB. Does it make a difference which one we use?

One reason to think that it might is that the two color spaces have different white points. Adobe RGB has the white point of the D65 Illuminant seen by the 1931 two-degree Standard Observer. ProPhoto RGB has the white point of the D50 Illuminant seen by the same observer.

My analysis program tries to level the playing field for files with different white points by translating the reference image under the reference illuminant into a color space with the same white point as the input color space, using the Bradford algorithm to effect the adaptation. If the raw developer uses exactly the same algorithm, then the white point of the raw developer export color space shouldn’t matter.

I exported the same file (a7RII raw of ColorChecker, default settings except for white balance, Adobe Standard profile, export sharpening off) twice from Lightroom: once in Adobe RGB and once in ProPhoto. Then I turned the analyzer loose on them. Here are the aggregate statistics that resulted:

pprgb cs argb stats

As you can see, the color space makes no material difference. I’m surprised how little difference it makes. This may mean that Lightroom used Bradford for adaptation.

Capture 1 vs Lr ASP default color processing

This is the 32nd in a series of posts on color reproduction. The series starts here.

It’s been a long, long road, but I’m finally ready to address the question that started this serial discourse on color reproduction, to wit: “What are the differences between Capture 1 and Lightroom default color conversions of Sony a7RII files?”

If you’ve read the studies and explanations that preceded this post, you won’t expect a simple answer. If you haven’t, what follows may confuse you. I will attempt to explain a few things as I go along, but I encourage you to follow the link at the beginning of this post and at least skim the material.

The test conditions are as follows:

  • Sony a7RII
  • Sony 90mm f/2.8 FE macro
  • Two Westcott bicolor LED panels, set to 5000K and full power
  • ISO 100
  • The C1 free version that only supports Sony cameras, default profile, default control settings
  • Lr CC 1015 current version, default profile (Adobe Standard), default control settings
  • Both raw processors white balanced to the third gray patch from the right.
  • Files exported from both raw processors in Adobe RGB.
  • Reference is illuminated in (simulated) D50 light.
  • Color adaptation with respect to the D65 white point of Adobe RGB performed using Bradford.

The overall statistics, with the parts that I consider to be worthy of special discussion highlighted in green.

Lr ASP vs C1 stats

The first thing to notice — the first row with green highlighting —  is that the overall accuracy of the two raw developments are quite similar. C1 wins by a nose on Lab 2000 Delta E, and loses by a slightly greater amount on plain old Lab Delta E and a bit more than that on CIELuv Delta E.

The next is that the C1 image is substantially more chromatic than the Lr one. Both processors err in the aggregate on the side of oversaturation, as indicated by the mean non-gray CIELab Delta Cab and Cuv rows. These are CIELab and CIELuv chroma errors; a positive number is oversaturated, and a negative number means undersaturation. The Lr average oversaturation is modest, and the C1 number is substantially higher, though not extreme.

Neither raw processor is linear:

a7RII C1 5000K Ill D50 Ref Bradford21

a7RII Lr CC Adobe Standard 5000K Ill D50 Ref Bradford21

Capture One is on top, with the blue line being the measured luminance readings of the gray patch row versus the correct ones, and the red line being a linear response. Lr is below, with a very similar curve that boosts midrange contrast, depresses shadows, and rolls off highlights. In Lightroom, this behavior comes from the Adobe Standard profile. I don’t know enough about C1 to tell you what part of the program creates this effect. It is almost certainly deliberate on the part of the designers of the two software packages, and it causes the color errors to be larger than they otherwise would be, but probably makes typical images look better to most people.

Now we turn to which color patches have the largest departures from accuracy, as measured in CIELab Delta 2000.

a7RII C1 5000K Ill D50 Ref Bradford7

a7RII Lr CC Adobe Standard 5000K Ill D50 Ref Bradford7

The differences are not striking, and the above plots are only likely to be useful if you have particular colors whose accuracy is important to you.

Now let’s consider chromaticity differences. First, we’ll look at CIE u’v’.

a7RII C1 5000K Ill D50 Ref Bradford26

It’s a little hard to see, so we’ll zoom in:

a7RII C1 5000K Ill D50 Ref Bradford27

The reference values are the open circles. The squares are the C1 values, and the diamonds the Lr ones.

Take a look at the two bottom triplets, which correspond to the two blue patches. In the case of both C1 and Lr, the rendition is away from magenta, and towards cyan. C1 oversaturates both more than Lr, as is indicated by radial displacement from the white point.

Looking at the same data in the a*b* (chromaticity) plane of CIELab:

a7RII C1 5000K Ill D50 Ref Bradford29

The same management of the blues is apparent, as is C1’s greater shading of Caucasian skin tones in the red direction. C1 also makes the bright yellow patch a great deal more saturated than a strictly accurate interpretation. A C1 shift of the orange patch in the yellow direction is also striking.

And lastly, for those of you out there who speak CIELuv, the same chart in u*v*:

a7RII C1 5000K Ill D50 Ref Bradford30

And here’s the big question to those who wondered about the differences between C1 and Lr default processing: does the above data help, or just confuse?

Macbeth testing with auto-exposure

This is the 31st in a series of posts on color reproduction. The series starts here.

We have seen that, with nonlinear profiles, Macbeth color accuracy can be affected by exposure. The purpose of this post is to explore the utility of the automatic exposure systems that are built into modern cameras as a tool to minimize these variations.

I took a Sony a7RII, and set it in aperture mode so that it would vary the exposure as I varied the light level. The lens was the Sony 90mm f/2.8 FE macro. The lighting was two Westcott LED panels on full, with the color temperature set to 5000K. I made 11 exposures. The first exposure was with the light intensity set to 100, then 95, then 90, then 85, then 80, and so on, all the way to 50, for roughly 1 stop of dynamic range in tenth-stop intervals. The intervals are actually linear, not logarithmic, but that’s not important for this test.

The simulated reference was lit with D50 light. I developed in Lr with the distinctly nonlinear Adobe Standard profile, and all controls at their default settings except that I white balanced to the third gray patch from the left. I computed the mean and standard deviation (sigma) of a bunch of aggregate color measures.

Here are the resulting statistics:

50-100 ASP A mode stats

These results are incredibly precise, with standard deviations (sigmas) two orders of magnitude below the means. It makes me think that the a7RII, although it reports shutter speed in 1/3 stop intervals when it’s in A mode, actually adjusts the shutter more precisely than that. [Edit: I have looked at the gray row values in Lr of the 11 files. The readings for the lightest patch are all withing a few tenths of the nominal Lr value of 89.2. That means that the a7RII metering is much more precise than 1/3 stop. Pretty impressive!]

It’s clear to me that the automatic exposure system in the a7RII can easily deal with the exposure requirements for consistent Macbeth chart results.

Colors in the real world? That’s another story, as anyone who’s tried to master the intricacies of exposure in situations where you’ve got to be bang-on in the camera, like with ‘chrome films. However, that’s really a different issue.

Macbeth light dimming effects

This is the 30th in a series of posts on color reproduction. The series starts here.

In some of the testing I’ve been doing, I’ve been varying the intensity of the Westcott LED panels that I’ve been using, and expecting the spectrum of the lights to not change materially. Is that a good assumption? I devised a fairly extreme test.

The camera was the Sony a7RII, set in aperture mode so that it would vary the exposure as I varied the light level. The lens was the Sony 90mm f/2.8 FE macro. The lighting was two Westcott LED panels on full, with the color temperature set to 5000K. I made seven exposures. The first exposure was with the light intensity set to 100, then 75, then 50, then 37, then 25, and so on, all the way to 3, for roughly 6 stops of dynamic range in half-stop intervals.

The simulated reference was lit with D50 light. I developed in Lr with Adobe Standard profile, and all controls at their default settings except that I white balanced to the third gray patch from the left. I computed the mean and standard deviation (sigma) of a bunch of aggregate color measures.

Here are the resulting statistics:

delta e vs light level stats

Not too bad, but a long way from great.

Let’s look at one of the measures, CIELab Delta E, versus light level setting:

delta e vs light level

You can see that things are pretty good until the setting is about 9, or a little over three stops down, then there’s a shift. Since we haven’t been using any levels below 50 so far, the work that I’ve already done should hold up. I’ll be careful in the future of dimming the panels below 10.

There’s one thing that I should mention. I ran the above tests during the day, and a little daylight leaked in to the room. It’s possible that the panels are doing fine, and that’s the reason for the shift.

I note that I white balanced the whole series to the exposure made with the light the brightest.

 

Macbeth ISO effects

This is the 29th in a series of posts on color reproduction. The series starts here.

Some have suggested making a different profile for every ISO setting that you use. Sounds like a lot of work to me. That’s undebatable. What is debatable is whether it’s necessary. I thought I’d find out, for one camera, anyway.

The camera was the Sony a7RII, set in aperture mode so that it would vary the exposure as I varied the ISO setting. The lens was the Sony 90mm f/2.8 FE macro. The lighting was two Westcott LED panels on full, with the color temperature set to 5000K. I made seven exposures. The first exposure was with the camera’s ISO “knob” set to 100, then 200, 400, 800, 1600, 3200, and 6400.

The simulated reference was lit with D50 light. I developed in Lr with Adobe Standard profile, and all controls at their default settings except that I white balanced to the third gray patch from the left. I computed the mean and standard deviation (sigma) of a bunch of aggregate color measures.

If you look at only the ISOs from 100 through 3200, it looks like ISO setting makes no difference:

Iso100-3200

The sigma, or standard deviation, values are very low. However, if we add ISO 6400 into the mix, the picture changes.

Iso100-6400

Still not bad, but it is clear that the ISO 6400 numbers are different from all the other ISOs.

If we look at the details, we can see the effect clearly:

ISO details

Whether this is the result of the camera doing something differently at ISO 3200 and ISO 6400, or whether it’s Lightroom and ASP doing something differently is not a question to which I presently have an answer.

It looks like you don’t need to make a profile for each ISO with this camera as long as you don’t care that much about accurate color for ISO settings above 3200. I’m certainly in that camp. You’ll have to decide for yourself.

With other cameras, who knows? More testing seems to be in order.

Photography meets digital computer technology. Photography wins — most of the time.

Entries RSS
Comments RSS