Kalles Fraktaler 2.13

  • 90 Replies
  • 4954 Views

0 Members and 1 Guest are viewing this topic.

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #75 on: September 06, 2018, 06:10:10 PM »
Would it be possible to add the following formulas? a and b could be taken from the Re and Im part of "factor a for Redshifter.

 z := ( x^2-y^2 + i (ax^2+2bxy) ) + c
 z := ( x^2+y^2 + i (ax^2+2bxy) ) + c

It's the most general (up to conjugation) quadratic form in 2D.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #76 on: September 06, 2018, 06:18:04 PM »
kf-2.13.9 released https://mathr.co.uk/kf/kf.html#kf-2.13.9

  -  new feature: auto skew without miniset: new button in Newton-Raphson zooming dialog, "Auto Skew (Escape)", that skews at the center pixels (algorithm suggested by gerrit)
  -  bugfix: set window title on parameter drag and drop (reported by gerrit)
  -  bugfix: reading PNG metadata works even if it is moved after the image data chunks and has a miscapitalized "Comment" keyword
  -  bugfix: fix Mandelbar derivative calculations for ADE (reported by gerrit)
  -  bugfix: fix 4th False Quasi formulas Newton-Raphson zooming

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #77 on: September 06, 2018, 11:32:58 PM »
That was quick. Mbar does not seem quite right though.
Attached location and result of skew from KF and what I get (using image center).


Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #78 on: September 07, 2018, 01:09:51 AM »
B-ship looks better though still quite different from what I got, but not worse.
Attached my code, slightly modified to run in Octave, though it's 2000 times slower in Octave than in MATLAB.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #79 on: September 10, 2018, 10:19:20 PM »
Mbar does not seem quite right though.

Looks ok (unskewed) when I zoom in (using the Page Up key).

Note that the auto skew escape is at pixel floor(width/2) floor(height/2) which is not quite at the image center when the dimensions are even, as they usually are... this could be the cause of the difference in this example.

Another difference is that the auto skew doesn't properly scale/zoom after the stretch: KF has a skew matrix like (s, 0; 0, 1) while the polar decomposition has (sqrt(s), 0, 0, 1/sqrt(s)), which is a factor of sqrt(s) different.  I just didn't get around to fixing it yet.

I haven't checked your octave code yet, but thanks for that.

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #80 on: September 11, 2018, 12:51:13 AM »
OK I see now, it just zooms out. Maybe better to skew where you click, like NR?

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #81 on: September 17, 2018, 12:46:48 AM »
I thought exif data was destroyed in photoshop, but apparently not. However, when I open (for example) this jpg (postprocessed in photoshop) in an exif viewer all the kfr data is there, but it does not open in KF: https://fractalforums.org/image-threads/25/gerrit-images/565/msg9402#msg9402.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #82 on: September 17, 2018, 02:49:48 PM »
I thought exif data was destroyed in photoshop, but apparently not. However, when I open (for example) this jpg (postprocessed in photoshop) in an exif viewer all the kfr data is there, but it does not open in KF: https://fractalforums.org/image-threads/25/gerrit-images/565/msg9402#msg9402.

exiftool -X shows the data of your image is stored in these EXIF tags:
Code: [Select]
xmlns:IFD0='http://ns.exiftool.ca/EXIF/IFD0/1.0/'  <IFD0:ImageDescription>
xmlns:IPTC='http://ns.exiftool.ca/IPTC/IPTC/1.0/'  <IPTC:Caption-Abstract>
xmlns:XMP-dc='http://ns.exiftool.ca/XMP/XMP-dc/1.0/'  <XMP-dc:Description>
but EXIF tags are not the same thing as JPEG comments (EXIF is more structured and modern, while JPEG comments are kinda legacy I think)

I'll add to my notes a feature request for KF to support EXIF reading.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #83 on: September 17, 2018, 03:00:52 PM »
Would it be possible to add the following formulas? a and b could be taken from the Re and Im part of "factor a for Redshifter.

 z := ( x^2-y^2 + i (ax^2+2bxy) ) + c
 z := ( x^2+y^2 + i (ax^2+2bxy) ) + c

It's the most general (up to conjugation) quadratic form in 2D.

I added these in my local test versions, but I swapped a and b (so the RSR factorA = b + a i) to make the default factorA=1+0i be more useful for this formula.

What should the name in the dropdown be?  General Quadratic Minus and General Quadratic Plus are my working titles so far..

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #84 on: September 17, 2018, 07:47:08 PM »
I added these in my local test versions, but I swapped a and b (so the RSR factorA = b + a i) to make the default factorA=1+0i be more useful for this formula.

What should the name in the dropdown be?  General Quadratic Minus and General Quadratic Plus are my working titles so far..
Nice. Names sound good. Exif reading would be nice. I bet there are thousands of post processed images around with kfr data in exif format.

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #85 on: September 19, 2018, 03:51:18 AM »
Background: I usually render at say 20000^2, save kfb, then tune (or completely change) coloring.
Recent parallelization improved response but it's still pretty slow, so I downsample (typically factor 10) the kfb.

Any ideas on how to best do this? For iteration coloring anything works, but for DE bicubic downsampling does not work well (result has completely different colors from the full resolution).  Best results I've gotten by taking exp() of the downsampled log() of the ADE data, but there are still color changes.

Matlab/Octave code (change extension to .m) attached.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #86 on: September 19, 2018, 02:53:54 PM »
maybe instead of resizing, you could crop various parts of interest and combine them into a new mangled kfb?  don't know a good solution, I think the geometric mean you are doing is probably as good as it gets?  or for ADE you could simply pick one sample at random from each 10x10 cell (supposing you are downsampling 10x)?

Offline gerrit

  • *
  • 3f
  • ******
  • Posts: 2265
« Reply #87 on: September 19, 2018, 06:49:32 PM »
OK, thanks. I thought maybe you had it all figured out. I was hoping for a way of downsampling kfb so rendering it would be identical to rendering the full kfb and downsampling the image, but that is not possible in general I think.

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #88 on: October 23, 2018, 08:20:40 AM »
kf-2.13.10 released https://mathr.co.uk/kf/kf.html#kf-2.13.10 bugfixes

- bugfix: corrupt image at transition between number types (eg near e600) (reported by CFJH)
- bugfix: changing "threads per CPU" during rendering crashes (reported by CFJH) (the menu is now disabled during rendering)
- bugfix: set approx terms to 3 for other than power 2 Mandelbrot
- bugfix: fix floatexp toString for negative values and 0
- internal: update to et-kf API version 5

Offline claude

  • *
  • 3f
  • ******
  • Posts: 1684
    • mathr.co.uk
« Reply #89 on: October 29, 2018, 08:46:37 PM »
kf-2.13.11 released https://mathr.co.uk/kf/kf.html#kf-2.13.11 bugfix release for the stable branch

  -  bugfix: fix translation with reuse reference enabled (reported by Dinkydau)
  -  bugfix: don't add additional references if autosolve is disabled
  -  bugfix: ghc-8.6 compatibility for preprocessor (MonadFail)
  -  upgrade to libpng 1.6.35


xx
Kalles Fraktaler 2 + GMP

Started by claude on Kalles Fraktaler

232 Replies
12664 Views
Last post September 12, 2019, 10:27:53 PM
by claude
xx
Kalles Fraktaler 2.14

Started by claude on Kalles Fraktaler

287 Replies
10483 Views
Last post Yesterday at 07:52:15 PM
by claude
xx
Kalles Fraktaler 2.15

Started by claude on Kalles Fraktaler

45 Replies
1012 Views
Last post Yesterday at 07:54:34 PM
by claude
xx
Another version of Kalles Fraktaler?

Started by greentexas on Kalles Fraktaler

9 Replies
1200 Views
Last post September 22, 2017, 02:59:16 PM
by greentexas
xx
The history of Kalles Fraktaler - comparison between old and new

Started by FractalAlex on Kalles Fraktaler

10 Replies
241 Views
Last post June 11, 2020, 03:19:33 AM
by FractalAlex