experimental imagery

Post your non-gallery photos here.
Post Reply
robertofbaycot
Posts:51
Joined:Mon Jan 30, 2017 7:25 pm
experimental imagery

Post by robertofbaycot » Sat May 02, 2020 1:13 pm

Wow! You people have been busy! Outstanding blackbird shots Lisa!

I've been experimenting with colour, based on the channel mixing approach in, for example, Affinity Photo.

Here are a couple of GIFs cycling through colour changes. (You may have to wait for half a minute or so before the GIFs start cycling).
fruit_rgbr.gif
fruit_rgbr.gif (9.4MiB)Viewed 1895 times
tangle_rbgr.gif
tangle_rbgr.gif (18.98MiB)Viewed 1882 times

Ed Jeffries
Posts:107
Joined:Sat Mar 14, 2020 11:20 am

Re: experimental imagery

Post by Ed Jeffries » Sat May 02, 2020 3:26 pm

Hi Robert - those GIFs are really effective with very seamless transitions. Really nice.

As a matter of interest, what machine are you running A.P. on? I've recently purchased A.P. and I find my m/c runs very slowly and sometimes hesitates. My m/c is a 2-core Lenovo running 4GB of RAM with a clock speed of 2GhZ. Do you think that is sufficient or should I increase the memory - maximum I can install is 16GB. It may also be that the overall spec of the machine is not sufficient to run that level of software.

I'd be very grateful for your comments and suggestions. Thanks.

robertofbaycot
Posts:51
Joined:Mon Jan 30, 2017 7:25 pm

Re: experimental imagery

Post by robertofbaycot » Sun May 03, 2020 9:21 am

Hi Ed,
I'm running AP on quite an old machine, so it's spec is nothing wonderful. It has 8 GB RAM, which is the most the machine can have. My immediate reaction to your post is that more RAM should have a dramatic effect on your machine's performance. I find that background processes alone on my Windows 10 setup can use nearly 4 GB. My processor is a twin core Intel Core i5 (at least 2 generations out of date!) running at 2.4 GHz. An SSD also makes an enormous difference. So, get some more memory and see how you get on!
Best wishes
Robert

Ed Jeffries
Posts:107
Joined:Sat Mar 14, 2020 11:20 am

Re: experimental imagery

Post by Ed Jeffries » Sun May 03, 2020 1:36 pm

Hi Robert - Thank you very much for getting back to me. It sounds that increasing the memory capacity is the direction to go in the first instance. You are right about the utilisation of memory when Windows 10 is operating in that I have discovered that I only have 1.92GB for actual use. No wonder the system fell off the cliff! I normally use Nikon Capture nx-d which works well but is limited in it's features (no layering, etc). I will certainly upgrade to the maximum of 16GB to ensure I have the capacity for the future. Once again, thanks for the very useful advice.

Keep the excellent images coming - you and the other members are all very accomplished photographers.

robertofbaycot
Posts:51
Joined:Mon Jan 30, 2017 7:25 pm

Re: experimental imagery

Post by robertofbaycot » Mon May 04, 2020 1:45 pm

The colours are a bit weird arent they?

The colour changes are not in any way random or curated by me;
they arise from a bit of mathematics that I've done and enjoyed enormously.
The colour changes are based on the idea of swapping the contents of
the red-green-blue colour channels in a colour picture, so,
for example, if you swap the contents of the red channel with the blue channel,
in a photo, red tomatoes will look blue, and something blue will look red.
Swapping the colours in this way is an extreme; I was looking for a way to get from
no colour change to a large colour change in small steps.

A colour change can be represented by a 3x3 table (a matrix) which is applied to
the original photo to achieve the colour change. You can then apply a
second matrix to get a further color change, and so on for as many changes as
you like. The colour changes shown in the GIFs arise from a single transformation matrix
applied again and again until you arrive back at the initial image.

The matrix for this example is:

0.985432 0.127322 -0.112754
-0.112754 0.985432 0.127322
0.127322 -0.112754 0.985432

(If you have a photo-editing program such as Affinity Photo, you
can try this out yourself.)

There are only 5 basic sequences that can be generated in this way.

Whether the colour changes are aesthetically pleasing is something else...

Ed Jeffries
Posts:107
Joined:Sat Mar 14, 2020 11:20 am

Re: experimental imagery

Post by Ed Jeffries » Mon May 04, 2020 4:21 pm

Robert, That sounds fantastic. Is the matrix manipulator part of the tools available in A.F. or is it something external? How are the elements in the matrix calculated? I haven't done very much with A.F. yet due to stability problems we now realise are due to insufficient memory. I need to upgrade so I can start to explore A.F. and extend my skill set further. I'll keep you posted on how I get on - need to get the dram packs first though.

robertofbaycot
Posts:51
Joined:Mon Jan 30, 2017 7:25 pm

Re: experimental imagery

Post by robertofbaycot » Tue May 05, 2020 11:09 am

Hi Ed,

There is a tool called 'Channel Mixer' in AF under their Adjustments tab. The elements of the matrix are entered using slidebars or numbers, though the accuracy with which you can enter values seems quite limited. There is an excellent macro facility in AF, so if you record setting up the channel mixer, you only have to do it once, then you can apply the macro. You can also apply it repeatedly. Be interested to see how well this works - I've used something else (see below) for the calculations.

The matrix is calculated as follows. Looking at channel mixing, the extreme cases involve interchanging whole channels, so, for example, the following matrix

R G B
R 0 1 0
G 1 0 0
B 0 0 1

swaps the red and green channels but keeps the blue channel in place. There are only 5 permutation matrices of this sort. What I wanted to do was to get to this transformation in a number of small steps, so I could see the colour variations along the way. Let's say I want 10 steps. So, I need a matrix, which, when multiplied by itself 10 times, gives me the matrix above. In other words, I need the 10th root of this matrix. There are matrix equivalents of exponential and logarithm functions, and the 10th root can be calculated using these as exp(log(M)/10). This is a not a simple calculation - you need a maths computer library to do it. I use numpy in Python. In the end, I do find it quite remarkable that a single matrix, containing a pattern of 3 numbers, can generate such a huge variation in colour. I've never seen a tomato that blue!

Ed Jeffries
Posts:107
Joined:Sat Mar 14, 2020 11:20 am

Re: experimental imagery

Post by Ed Jeffries » Wed May 06, 2020 12:18 am

Good Evening Robert - well, you certainly are a very technically adept photographer. I'm impressed that you're also a Python coder. I have only done some 6502 microprocessor assembler and compiled PC Basic in the past at Chelmer Institute - nothing to what you are doing though.

I was looking into the colour mixer and saw that there was a LUT - is that the one where you put the matrix elements? I must get the additional memory so I can start using A.F. Well done - keep the great images coming.

robertofbaycot
Posts:51
Joined:Mon Jan 30, 2017 7:25 pm

Re: experimental imagery

Post by robertofbaycot » Sat May 09, 2020 11:25 am

Hi Ed,
I think LUT is a separate tool - not sure what it does. The tool I was talking about is called 'Channel Mixer' & is in the same Adjustment group as LUT.

I worked in electronics research for 40 odd years, so I'm very techie! Python is a much nicer language to use than either of those you mention - it's very powerful too - it only needs one line of Python to apply one of these matrices to an image. Mention of assembler brings back some ancient memories! What's your background?

Have you got your RAM yet?

Ed Jeffries
Posts:107
Joined:Sat Mar 14, 2020 11:20 am

Re: experimental imagery

Post by Ed Jeffries » Sat May 09, 2020 3:49 pm

Hi Robert, I also have a background in electronics but not so much in design as you obviously have but in installation, operation and maintenance. Worked over the years on Troposcatter comms links in the Far East, pilot training systems at Essex Uni and security systems in South Africa. Was also Technical Manager for Hermes Precisa (Olivetti) for 5 years. A real varied work history - picked up an HND along the way. Now I'm taking photographs! Who would have believed that!

Post Reply