Mike Chaney's Tech Corner

Technical Discussions => Articles => Topic started by: admin on March 02, 2015, 07:43:24 PM

Title: March 2015: That Blue/Black Dress. Or is it White/Gold?
Post by: admin on March 02, 2015, 07:43:24 PM
This Blue/Black and White/Gold Debate Can't be Real, Can it?

I have to admit, I'm a little late to the party on this one.  By social media standards, several days is an eternity and this dress with the black and blue or white and gold stripes has already been beaten to death and many have moved on to this week's craze (whatever that might be).  Still, the reason I'm a bit late to chime in on this one is that when my kids first showed me the tumblr post with that dress last week, I thought it was just another internet spoof designed to grab attention.  I thought to myself, no way is anyone going to see white and gold in that dress: this is nothing more than the old foolish "made you look" trick.

I was floored when I saw it yesterday on the front page of CNN.  Are some people really seeing white and gold, I asked myself?  I mean... people who are not color blind or suffer from some other eye ailment?  Apparently so.  I started digging.  I asked our own Fred from the forums here: he saw white and gold.  WHAT?  I almost fell out of my chair!  While the majority seems to be in the blue and black camp, some polls are showing at least 25% are adamant that the dress is white and gold!  At this point, I know you've seen it, but I thought it worth digging up the actual original image that started the debate because by now there are thousands of copies all over the internet, many of which have been "sanitized" or had their color changed in some way.  So here's a link to the actual image on tumblr that started it all.  Hopefully it'll be around for a while: http://swiked.tumblr.com/image/112073818575

What color is the dress?

OK, so you've seen it.  Now, if there's one area where everyone seems almost 100% united, it's in wondering "how can anyone see almost the opposite of the colors I see?".  Other than a handful of people who claim to be able to see both, if you see white and gold, you'll always see white and gold.  And if you see blue and black, you'll always see blue and black.  So what makes us see differently?  Are some people's eyes defective in some way, such as having a blue deficiency?  Or are all eyes the same and some brains are just wired differently?  I'm no expert in biology or minds, but I do have decades of experience in digital photography where I've written close to a million lines of code designed to push pixels around and get the correct color to your eyes!  So let's take a look at what we know, from the perspective of the digital image itself.  I'm not going to delve into the technical aspects like where you got the image, whether or not your monitor is calibrated/profiles, whether you are looking at the image on a monitor or a phone, and so on.  Let's assume you are looking at the original image on a device that can display web images reasonably accurately.  I think that covers most devices in this case because we're really not splitting hairs with this topic.  Meaning that it seems you can vary the display device, brightness, and most other parameters within reason, and blue/black people will still see blue/black while white/gold people will still see white/gold.

Let's start with a loaded question, shall we?  "What color is this dress"?  I could easily measure the colors on the dress and tell you that the colors coming off the screen are Wild Blue Yonder and Makura (derivations of blue and brown), and you might be tempted to tell team white and gold to go home.  But "what colors are coming off your screen" was not the question!  As soon as you are asked to evaluate what color the dress might be, you are being asked to interpret the image on the screen.  When I first saw it, I said blue and black, but even that was an interpretation.  If someone had asked, "What colors do you see in this image" (pointing to the area of the dress), I would have answered differently and said "light blue and dark brown" or just "blue and brown".  But my mind interprets the brown as black because of experience: the bottom dark stripes are much darker than the top and the darker ones are close to black... and "who would make a dress with a gradient of mud brown stripes that vary from top to bottom?"  So my brain tells me it's probably blue and black in reality.  My experience in digital photography sways my opinion as well.  As soon as I see the image, I can tell it is overexposed, so my mind has already made the assumption that those dark brown stripes are probably really black but just photographed too bright so they end up a little brighter than black with a bit of a reddish tint: brown.

But what about white and gold?  I've read that some people think it's just a difference in how people interpret different lighting and what they assume is white.  Or that the lighting in the picture isn't good which makes everything up for interpretation.  But I think it's more complex than that.  Technically, the brighter stripes around the middle of the dress are a color called Wild Blue Yonder (about RGB 125,133,180).  Of course, it's going to differ depending on where you decide to measure, but that'll be close enough for our purposes.  Wild Blue Yonder is akin to a dull sky blue.  Like a blue sky on a hazy day.  I must say, it's nowhere near white!  To my eyes and mind, it's too blue to be just bad white balance: we've all seen pictures that have a blue tint and we know the shot has the wrong white balance.  In that circumstance, white stripes would technically be blue, but we would interpret them as white based on other reference points in the image and what we expect (like skin tones that we see are also too blue).  But the bright stripes in the dress are not just a slight blue "tint".  They are blue.  I mean real blue: too far off from white for most people to deem just a "slight blue tint"!  So why do some see white?

Nature or Nurture

Our eyes adjust for lighting all the time.  A white piece of paper looks white under an incandescent lamp, sunlight, and sodium street lights.  Our eyes and brain work together to readjust "what is white" in those situations.  Once a good white (or gray) balance is known, the rest of the colors fall into place.  But at this point we have to ask, how far are your eyes/mind willing to go to adjust for what is white?  I suspect like many things, there's a bit of both nature and nurture at play here.  Some people may have more sensitive blue cones in their eyes.  Some may have minds that are willing to adapt farther to "redefine" those blue stripes in the dress to white based on conditions you are exposed to in daily life.  And some may have a greater range of light adaption (basically dynamic range): simply put, due to the physical qualities of the eye like the rods, cones, and how your own pupil dilates, you may see some images brighter or darker than average.  If your eyes/mind interpret the image significantly brighter than most, that could be all it takes to swing you from blue and black to white and gold: see below!

An Illustration

Take a look at the illustration below.  It shows the exact same dress on the left/right and a single color stripe on the top/bottom.  The top color stripe is the same 125,133,180 RGB all the way across, and the bottom stripe is 107,93,79 RGB all the way across:


I cut out the original dress and placed it on a gradient: it is the same dress with the same colors on the left and right.  Notice how the dark stripes in the dress on the right look closer to black while the same stripes on the dress on the left look closer to gold.  The bright stripes in the dress also look "bluer" in the dress on the right and more faded on the dress on the left.  So your brain is already playing tricks on you depending on whether it interprets the dress as being in shadow or bright light.  I can assure you, the two dresses are identical on both sides.  Now take a look at the bright and dark stripes on the top and bottom.  These are the colors of the stripes in the dress at about the middle of the dress.  And no, those wide rectangles of color on the top and bottom are not gradients (shaded).  They are exactly the same color all the way across!  Your brain interprets the one on top as a lighter blue on the left and a darker blue on the right.  It also interprets the bottom as closer to a very dark brown on the right and more of a gold on the left.  At least with this illustration, maybe you can begin to see where "the other" camp is coming from.  The illustration shows the concept; the rest is just about extremes.


I wouldn't be surprised to find that there are subtle differences in color response of different eyes; something that goes beyond the typical crude color blind tests.  Maybe the white and gold camp has a slight blue deficiency?  Maybe the blue and black camp has a red/green over-sensitivity?  What is right?  What is wrong?  Who knows: maybe this has opened up a whole new type of evaluation for human color vision or might give us a new idea on how to test the differences.  One thing I did find interesting is that I asked Fred to print the photo and look at a print instead of a monitor.  Given that he is in the white/gold camp, I wanted to see what difference that made since a print sends information to your eyes via reflected right.  He replied that instead of white/gold, the print looked blue/brown.  So apparently there is some difference there, as I thought there might be.  Try printing the photo and see if you change camps!

And as long as you're here, if you are interested in the digital photography aspect of this discussion, you might find my Qimage Ultimate professional photographic printing software (Windows) handy: http://www.ddisoftware.com/qimage-u.  It'll make sure all your devices are delivering accurate color that matches real life.  How you interpret those colors... that's up to you!  ;)


Title: Re: March 2015: That Blue/Black Dress. Or is it White/Gold?
Post by: Terry-M on March 02, 2015, 10:56:00 PM
Hi Mike,
Thanks for your interesting article on this. It was all over the UK  BBC news at the weekend.
The BBC also talked about light quality and demonstrated how exposure and the ambient light colour temperature could affect the acual colours in the image. They also showed the actual dresses in the shop (they were then selling like mad by then) and they are a mid blue and black. I'm sure BBC TV get their colours right!
The original image that was shown seemed to me to be overexposed and taken in warm artificial light.
I did a few tweaks and got it somewhere near ( the blue is probably too bright) to what I remember from the dresses shown the shop. See attached.

Title: Re: March 2015: That Blue/Black Dress. Or is it White/Gold?
Post by: admin on March 03, 2015, 01:11:21 AM
Yes, as of yesterday, sales of that dress were up over 800%.  And people are now selling "Team Blue/Black" and "Team White/Gold" type t-shirts.  :)

The actual dress is close to a royal blue and black, so the overexposure is what sparked this debate for sure!  It still blows my mind that anyone can see the brighter stripes as anything but blue, but when I showed Fred and gave him the two choices, he immediately said white and gold, or at least white and some sort of "mud" color.  If I pushed him and he had to revise it, he'd only go as far as silver and gold.  For me personally, I tried everything...  I blocked off everything but the dress.  Tried staring at a full brightness red screen first to adapt my eyes differently and then switching to the dress.  Full blue screen, then switched.  I even gave the whole image a 50% boost in brightness.  The brighter stripes still look blue.  Light blue.  But obvious blue to me.  The original is certainly not royal blue (due to the poor exposure) but it's a "blue enough blue" that it's almost blue sky color to me.  Certainly way too blue to just be a caste on the image that my mind might try to remove automatically.

I'm curious what you see?  Speaking of the original image.

I still think there may be something going on from a physiological perspective here.  Do some people live or work in such varying lighting that some people's brains have a wide swinging "auto white balance" correction in their brains?  Is there some variation in the rods/cones that no one realized was this common?  Again, I don't like any of the "answers" I've seen.  One article I read suggested that people see it differently based upon whether they are assuming the dress is in bright light or in shadow.  I think my illustration above disproves that: because while the colors do appear different (even though they are not), I still see blue/black on both sides.  Or to be technically correct, blue and brown.  But certainly not anything mixed anything close to white!  I find it hard to believe that there could be such a large swing just caused by the mind and how far their minds are willing to go to shift white balance in their brain either.  I suspect this all starts with a physiological difference in the actual "data" coming out of the optic nerves.  Perhaps there is a non-linear response at the bright/desaturated blue end of the spectrum where people have more varying response in that area than we once knew; based on rods/cones reacting more differently than we thought from one person to the next.


Title: Re: March 2015: That Blue/Black Dress. Or is it White/Gold?
Post by: Terry-M on March 03, 2015, 08:01:16 AM
Hi Mike,
I'm curious what you see?  Speaking of the original image
I saw pale blue and a dark brown in the original image.

Title: Re: March 2015: That Blue/Black Dress. Or is it White/Gold?
Post by: Jeff on March 03, 2015, 08:45:57 AM
Thanks for your dissertation on colours.  Very interesting.

To my eyes the blue and brown stripes do appear to change left to right, but the dress? my eyes see the right hand image just and only just slightly darker or say slightly more saturated certainly nothing like as much as the stripes. I presume that is my brain doing its job.  Good to know it's working :) :)

I have often thought I see colours differently to other people - my wife being one.  She often comments that my greens and yellows are too bright. I frequently readjust yellows and greens when I use default SelCol in Ultimate.

A couple of years ago I asked my optician if there was a test for colour perception.  All I got was a blank look, I wonder if there is anything in the way of charts like the standard eye test, say a  calibrated colour and alongside similar colours just slightly off the calibrated colour, but then probably our 'eye' would do all sorts of adjustments and render such a test useless.