Gryff: AKA Coontastic Gryffindor Update 25 May 2020

[Lulu22]

Just stunning

[richardandtracy]

It's lovely. I'm sure I've said before but you must be more than a little pleased with your charting program, with results like this.

Yes, I am happy with my charting program, but it is slowly highlighting a problem. I am using public domain values for the red, green & blue colour components of the DMC threads. As far as I can tell they trace back to an unofficial release by DMC (and it seems to have been made unofficial so they couldn't be held to it contractually). There seems to be a bit of a problem with the accuracy of these colour definitions. I know that sounds bad, but I'll try to explain what I think has happened.

Digital images can give component values for the red, green & blue components within the range 0 to 255, thus enabling 2^24 possible graduations of colour, or more different colours than most humans are able to resolve. It is defined that absolute black is 0/0/0 for Red, Green & Blue, while with brilliant white has all component values at 255/255/255. The problem comes in that real thread colours do not cover the whole range, because brilliant white still absorbs some incident light and does not reflect 100% of the light impinging on it. So the component values if you photograph brilliant white may be 239/239/239 (I have done this to see the numbers). Then, because black thread does reflect light, and not absorb every photon that impinges on it, the component values are 8/8/8. As a result the range is squashed a bit, covering 231 divisions of the possible 255. Seeing this, I think the originator at DMC shifted & stretched the colour values using a formula similar to that below:

• New Component Value = (Original component Value - 8 ) * 255/231

This then stretches the range to give black = 0/0/0 and brilliant white = 255/255/255. That enables all digital photos to be mapped easily against the thread colours.

BUT there is a problem with this simple idea.

The light end of the range is stretched more than the dark end. The dark end of the range is stretched 8 divisions out of 255, while the light end is stretched twice as far, meaning that on average the thread colours in the thread definition file will be lighter than the real threads. So, when you convert the colours, the stitched picture seems darker than the digital image.

I am not completely happy with this. I feel that there should be no compromise over the colours, and consequently the thread definition file value should be as close to the colour of the actual threads as it is possible to get. That way the mock-up of the converted image would be functionally identical to the stitched image. There are two consequences of this:

1. The bright sections of the photo may seem to be slightly washed out & lacking in detail, as there will be fewer thread colours in the brighter range of the colour component values and
2. I will have to purchase every thread colour, stitch a 10x10 block and photograph the whole piece with all 447 colours in one go before I can get sufficient real data to be 100% certain I have the best basis for my colour component list.

The big problem with this is the time it's going to take & the fact I don't yet have every colour.

I have done something similar with the Anchor Range of colours, in that I downloaded an official Coates pdf of the Anchor floss card that uses threads, and harvested the average colours off the card (using the part of my program where you can create a thread definition file). This then ensures the shadow cast by the thread is included in the average colour - which is also part of the colour you see in the cross stitch picture. I have not altered the Public Domain DMC colours, but I do now offer an optional adjusted DMC range that makes the thread definition component values 10% darker, which I think goes some way to solving the issues.

So to summarise all my waffle above:
I'm happy with the program. It is now sufficiently accurate in its conversion to highlight the relatively small errors introduced by assumptions made in generating the reference data. From now on, in order to make the program better, I am going to have to improve the quality of the reference data. And that's hard work.

Regards,

Richard.

[Purple Patsy]

As you make your prog freely available to all of us it would seem only fair that we all help you a bit with your research. If we all did a few 10 x 10 squares and sent them on to you (in an organised manner) that would help a bit surely? And it wouldn't take very long for each of us to do our bit.

[SteveM]

That is a very plausible explanation as to why stitching often turns out darker than expected. JoAnn fabric stores here in the USA are about to have a DMC sale of 4 skeins for $1 USD so I'll happily chip in some stitching patches and/or skeins.

[richardandtracy]

Thanks for your suggestions, I would like to think it over a bit.
I really don't want to put anyone to any expense, and don't want to put anyone out either. I feel it needs to be stitched on the same backing fabric so there is no colour cast introduced from that, and preferably on something like 18 count so the fabric shows through as little as possible. It would be a pain sending that around to people if they have a bit of a missing colour in their stash.

Maybe the best & quickest way would be for me to buy a DMC real thread card and have done with it. Our local LNS has changed hands recently and is now being run by a lady who attends the local knit & natter I go to every other week. Fortunately she doesn't know me well enough, so she won't hike the price up.

Anyway, I'll think on it a bit.

Regards,

Richard.

[richardandtracy]

This morning I ordered the DMC thread card from my local LNS.

May be here in a week or so.

Regards

Richard

[Serinde]

Thread cards are the best way to go, and can be a revelation -- or at least the ones I received from Renaissance Dyeing were, as they included your actual bits of wool. I immediately desired All The Colours Now.

[Mabel Figworthy]

I immediately desired All The Colours Now.

I do that anyway, proper shade card or no!

[richardandtracy]

I have a large (A4 + an inch all round) Ziploc document wallet I keep all my DMC skeins in, separated into smaller bags for 100-199/200-299 etc. I have fewer than half the range, and can't zip it up. Hrrumph. Goodness knows how big a bag I'll need when I'm closer to having the full range.

I'll keep going with Gryff until the snow above his tail is done, post a photo, then rotate to the Luz project for a month or so.

Regards,

Richard.

[wendywombat]

You obviously need more Ziplock bags!!

Looking forward to the update.

[RosemaryD]

Your post about DMC component values makes a lot of sense to me. The human eye/brain is surprisingly good at detecting small colour variations so I would imagine a relatively small inaccuracy might have a large apparent effect.

Even with the shade card that's going to be a lot of hard work but it will certainly be interesting to see the result.

[richardandtracy]

I have pottered for just over a month on Gryff now, progressing with page 11 of the chart. I have now finished all the background snow above his tail, as a result he's now going to make way for a few weeks work on 'Luz'.

Here's the chart cross reference illustration:


And the actual progress:

To the best of my knowledge, there are two stitches in the wrong place, where they are swapped over. Not that I could identify them without my marked up chart, now. The picture is coming on well, but it will be nice to see a bit more of the main subject.

The stitches are tiny, and it takes quite a bit of concentration to do them this small. I suspect that after Gryff & Luz are complete I'll take a break with an 18ct picture as it'll seem so large!

Regards,

Richard.

[Mabel Figworthy]

Oooh yes, I can see them - oh dear, they ruin the whole picture, those two switched stitches!

Or perhaps not.

I'm surprised you can identify them even with the chart !

[richardandtracy]

I circled them on the chart when I realised I'd put the dark grey blend in the wrong place so it wouldn't throw my counting when filling in with the dark blue blend. It's somewhere in the middle, a few rows below the centreline stiches.

Regards,

Richard.

[kingfisher68]

Awesome work done, lovely to see more contours of the tail & more ginger colours. Amazing how many there will be in the finished cat.

[SteveM]

Excellent work. Except for those two stitches of course.

[rcperryls]

Beautiful progress! Besides the glaring error of the 2 swapped stitches (which are totally not visible to anyone but you) this is going to be stunning!

Carole

[xstitchalanna]

He is coming along nicely! Such a pretty cat

[karen4bells]

It looks amazing!!! I definitely do NOT see any errant stitches!! Fantastic update!!!

[fccs]

Gryff!!!! I love seeing this one and now I have to be sad for awhile until he reappears. Your work is absolutely fantastic, two missed stitches included.