Kalles Fraktaler 2.9 and KFMM 1.30

[Kalles Fraktaler]

I have made another small update:

KF 2.9.1

• In this thread, http://www.fractalforums.com/mandel-machine/glitches-and-crashes/msg85895/#msg85895, I saw a nice suggestion of enhancing the exploring experience by not filling glitches with one colors, but instead let them remain the stretched pixels from the previous view. This can be disabled by the menu "Actions->Special->Show Glitches
• Enhanced Distance Estimation

KFMM 1.31

• Fixed color cycling when color division is more than 1. This is for old stylish solid coloring
• Enhanced Distance Estimation
• A second option for entering speed, "Time NS" = Time No Smooth. Because the smooth speed transitions is difficult to control

[claude]

I really like the exploration experience improvements (mouse wheel zoom around cursor location, using the previous view to give an idea of the zoomed view, animating the transition, only updating completed pixels).

I've not got the hang of the distance colouring though (I'm guessing it's calculated using a method similar to slopes rather than true derivative-based distance estimate?).  The palette repeats in a way that I don't like - I'd rather it stops at the end, so dense regions aren't so aliased and grainy while sparse regions are still bold.  I ended up combining 4 slope-colouring images at 90deg angle increments in GIMP using the Darken Only layer mode, then post-processing a bit (contrast stretch, blur to remove some grainyness, downscaling), to give something that looks more like what I want (top image, second image is one of the angles, parameter file attached too).

Some other minor things I ran into:
* auto-solve glitches isn't on by default, which meant that my first zooming ended up pixelated at double precision limit (I don't remember exactly which option fixed it, maybe it was a different option)
* reuse reference can't be selected at the same time as auto glitch solving
* I couldn't get the palette waves (the one below the infinite waves) to work properly, no matter what I chose for offsets it just ended up shades of grey when all the periods were the same
* reducing contrast then increasing it again gives a posterized effect (colour precision is lost when reducing contrast)
* distance colouring didn't do what I expected (see above, maybe an "all-directions/fake-de" toggle for the slope colouring would be an idea?)
* would be cool to be able to set the bright and dark colours for slope colouring, rather than just white/black - or even a second palette that gets applied according to the angle of the slope (which could be used for the previous point too)
* at one point the timer stopped updating in realtime, but when the image completed it showed the time (5:48.923 done ref 34)
* the menus are a bit confusing (reorganising them into submenus for different aspects could help)

I tested the 32bit version with WINE on Debian.

[Kalles Fraktaler]

I really like the exploration experience improvements (mouse wheel zoom around cursor location, using the previous view to give an idea of the zoomed view, animating the transition, only updating completed pixels).

Thanks

I've not got the hang of the distance colouring though (I'm guessing it's calculated using a method similar to slopes rather than true derivative-based distance estimate?).  The palette repeats in a way that I don't like - I'd rather it stops at the end, so dense regions aren't so aliased and grainy while sparse regions are still bold.  I ended up combining 4 slope-colouring images at 90deg angle increments in GIMP using the Darken Only layer mode, then post-processing a bit (contrast stretch, blur to remove some grainyness, downscaling), to give something that looks more like what I want (top image, second image is one of the angles, parameter file attached too).

Indeed, DE is inspired by Mighty Mandel
However I have attached what I wanted to achieve, several layes with different colors.
Background is blue. First layer is black, second layers is dark green and third layer is yellow. Each layer comes from a close passage of a Minibrot.

Some other minor things I ran into:
* auto-solve glitches isn't on by default, which meant that my first zooming ended up pixelated at double precision limit (I don't remember exactly which option fixed it, maybe it was a different option)
* reuse reference can't be selected at the same time as auto glitch solving

You cannot combine reuse reference and automatic glitch correction, since reuse reference means that one reference only is used and glitch correction involves several references. Reuse reference is meant to be used for the frames closest to the final minibrot in a Movie

* I couldn't get the palette waves (the one below the infinite waves) to work properly, no matter what I chose for offsets it just ended up shades of grey when all the periods were the same

If you use e.g. 4 periods on 128 colors, you would need to offset it with 16 in order to move it half a period.
If you use 32 periods on 128 colors, you would need to offset it with 2 in order to move it half a period.

* reducing contrast then increasing it again gives a posterized effect (colour precision is lost when reducing contrast)

Yes, reducing contrast to grayish colors and then back makes color information to be lost

I need to come back on the other comments

[claude]

However I have attached what I wanted to achieve, several layes with different colors.
Background is blue. First layer is black, second layers is dark green and third layer is yellow. Each layer comes from a close passage of a Minibrot.

Very nice!  I'll try the parameter file soon to see how you did it.

You cannot combine reuse reference and automatic glitch correction, since reuse reference means that one reference only is used and glitch correction involves several references. Reuse reference is meant to be used for the frames closest to the final minibrot in a Movie

I described a method for finding secondary references using the glitch detection test:  http://www.fractalforums.com/announcements-and-news/pertubation-theory-glitches-improvement/msg84154/#msg84154
The primary reference can be re-used (along with its series approximation coefficients), the secondary references don't need to be recomputed from scratch (they are found at their own period) - it turns out that a single iteration of Newton's method is almost always enough, and using the double precision derivative computed for distance estimation is usually enough, so it's quite cheap to find the next reference.  I haven't released my implementation yet, it still has too many annoying bugs (sometimes zooming too quickly in the GUI crashes in an infinite loop waiting for worker threads that already exited, sometimes a bad primary reference is used giving a blank screen or other wrong appearance...).  But no time to fix them for a few months...

If you use e.g. 4 periods on 128 colors, you would need to offset it with 16 in order to move it half a period. If you use 32 periods on 128 colors, you would need to offset it with 2 in order to move it half a period.

Thanks for the explanation   I thought period meant the length of the cycle, but it seems it's the number of cycles inside the palette length (so more like a frequency than a period).

[windwakr]

Hi, just want to say that this software is amazing! I never thought I'd be able to view deep zooms of the Mandelbrot set so quickly.


I have a request, though. Could you add a lossless image save option(like PNG or BMP)? JPG has too many ugly artifacts, even with high quality. I was looking into adding this for myself, but realized the source you provide is from an old version. I'd be missing out on newer features if I built onto that, so I have to ask you to add it.

[_BlueGuy_]

How do u save settings!?

[Kalles Fraktaler]

How do u save settings!?

Select "save".....

[_BlueGuy_]

but window size, pallete, etc

edit: figured it out