Kalles Fractaler and fancy render effects...

[stardust4ever]

I didn't want to bump a months old topic so I'm starting a new one. As some of you are aware, I've been in and out of the fractal scene for a while. I've seen some new render techniques employed in Kalles Fractaler both videos and stills and am at a loss on how to implement these effects. One thing I think is really cool, the distance estimation approach seen here:

I also wanted to implement Distance Estimation.
This is actually a cheat that is related to slopes, so no derivative is calculated because I don't know how to do derivative with perturbation
<a href="https://www.youtube.com/v/NKwkujUkMr0&rel=1&fs=1&hd=1" target="_blank">https://www.youtube.com/v/NKwkujUkMr0&rel=1&fs=1&hd=1</a>

All functions in the dialog boxes are now explained with tool-tips.
Since the test department is on holiday, there may be bugs

I think some of my "X" chromosome fractals would look cool rendered with this approach. On issue with color gradients, is that even with infinite waves or an attractive static pallet, the colors often don't line up with the features in such a way as to make them visible. This is especially a problem when rendering zoom movies, as I will pass through multiple features, and the color pallet always gets arbitrarily applied. I also toyed around with the idea of adding sloped effect to my videos.

I saw some of the "3D" videos rendered with MMY3D, and I gotta admit, not sure if I'm a fan of the "infinite bottomless pit" approach to 3D rendering, but I've seen in more recent videos and stills from Kalles Fraktaler, the use of distance approximation to generate a kind of bump map with which contrast is applied. This creates the illusion that the dendrites have a relief, similar to that of mountains. I like this. With regards to my Chromosome renders, I would like to use distance estimation to create the illusion that the features and spirals are "glowing" against a darker background. Perhaps even multiply the grayscale map with a rainbow style color pallet for neon effects. So often the details are either washed out and low contrast, or the image becomes way too "busy" so I set the periods of the color pallet or infinite waves to a happy medium that hopefully shows off detail without creating busy images.

I know with deeper zooms, this would be almost impossible without continually tweaking the color cycling along the way. it would be need if I could make stuff "glow like neon signage" or at least generate a relief map to create the illusion of texture. I hope this makes sense.

[Kalles Fraktaler]

I don't think any of us playing with perturbation has managed to calculate the derivative and perturbation, in order to create real distance estimation?

So the distance estimation is made based on the slopes.
Because of this, large antialias is necessary to produce good result, I used 3840x2160 for the movie you linked.
And also, for dense pattern the zoom size 2 of key frames is too large and produces visible squares, so I used 1.25, which produce more than 3 times of key frames.
Further, the palette is built with one color of background, white in the linked movie, and the rest of the colors black.
For other Movies with several layers, I used several large areas of the same colors, some entries with background colors in between.
Slopes and DE made in this way requires adjustments for every movie frame, and are very time consuming to render in KFMM.

I am sorry but it is not just a click of a button... 

[claude]

I don't think any of us playing with perturbation has managed to calculate the derivative and perturbation, in order to create real distance estimation?

In mandelbrot-perturbator I am using perturbation with true distance estimates.  The series approximation is mentioned here http://mathr.co.uk/blog/2016-03-06_simpler_series_approximation.html which I use to initialize the unperturbed derivative which is updated by the regular dc → 2 z dc + 1 iterations.  (Using perturbed iterations for derivative doesn't make too much sense because the derivative diverges quickly) I also use this derivative for one low-precision Newton step when finding secondary references at glitch detection time, to get a slightly better secondary reference.

[stardust4ever]

I don't think any of us playing with perturbation has managed to calculate the derivative and perturbation, in order to create real distance estimation?

So the distance estimation is made based on the slopes.
Because of this, large antialias is necessary to produce good result, I used 3840x2160 for the movie you linked.
And also, for dense pattern the zoom size 2 of key frames is too large and produces visible squares, so I used 1.25, which produce more than 3 times of key frames.
Further, the palette is built with one color of background, white in the linked movie, and the rest of the colors black.
For other Movies with several layers, I used several large areas of the same colors, some entries with background colors in between.
Slopes and DE made in this way requires adjustments for every movie frame, and are very time consuming to render in KFMM.

I am sorry but it is not just a click of a button...  

So this would probably be useless for extremely deep locations then. Still it would be fun to do stills in this manner. If I could get the "glowing dendrites" effect down that would be neat. Basically a simple brightness map based on change of neighboring pixels, or the bump map Kalles Fraktaler generates prior to applying the texturizer filter.

Anyways, I thought I would share this interesting grayscale result of a pure texturized fractal render over solid black. They look kinda like ice crystals...
http://stardust4ever.deviantart.com/art/Reactor-Core-Black-Texturized-597127758

Rendered as 12800x12800 grayscale, downscaled to 3200x3200 for antialiased effect.

This technique could be used to quite striking effect in low density areas thinks. If a "bump map" based on iteration density alone (used to generate the texture) could be exported to file, some really neat effects could be done in Photoshop or GIMP if combined with a normal color mapped iteration data of the same fractal.

[stardust4ever]

Another thing while I am on my soap box:

I rendered a 1080p 60Hz 14:25 video using slope and color cycling. The zoom movie (1289 3840x2160 frames to final depth of 3.00 E387) took approximately ~39 hours to render on my 8-core AMD Bulldozer @ 4.2Ghz. That's using H264 High profile "Slower" with Constant Rate Factor = 21. "Merge Count" = 4. Time to render AVI file was nearly as long as it took originally to render the original zoom frames. Just wanted to say however slow, the results are amazing! However I am currently doing a second pass at a slightly higher color cycle speed (0.10 instead of 0.05, another 39+ hours wait time) and will compare the two results before deciding which AVI file I will ultimately upload to Youtube.

Is there some reason why it takes such an extremely long time to render videos with color cycling enabled? With color cycling disabled, these 15-minute videos only take a couple hours. If code could somehow be optimized to render video frames more quickly, it would be appreciated.

I just wanted to add again, the result is truly spectacular! Slope rendering really make the dendrites pop out, and the amazing 3D slope effect gives definition to those areas which have iteration density too great to render the colors.

[quaz0r]

with this "keyframe" approach to producing animations, i guess he is just rescaling each rendered keyframe to produce each frame when there is no cycling or anything?  whereas with cycling each frame needs to be re-rendered since each frame is different, which i imagine he has "optimized" about as well as one can expect.

[stardust4ever]

with this "keyframe" approach to producing animations, i guess he is just rescaling each rendered keyframe to produce each frame when there is no cycling or anything?  whereas with cycling each frame needs to be re-rendered since each frame is different, which i imagine he has "optimized" about as well as one can expect.

It's not a big deal really. But AVI rendering is over ten times faster without color cycling. I imagine the Keyframe application generates a large 24-bit bitmapped image which is custom scaled to generate every frame. Because the pallet data is being rotated, it may be necessary to generate a new "bitmap" for each AVI frame, as opposed to once per zoom level. Since I am rendering a 60Hz video with 40 frames per zoom level, it has to render many, many more bitmaps. I am also utilizing the new "slope" feature which likely adds to the computation overload. I haven't tested if slope rendering is any faster without color cycling. I imagine if slope data were recomputed once every zoom frame as opposed to once every AVI frame, it might create noticeable "jumps" when the slope size suddenly doubles between zooms intervals.

Either way, I am very pleased with the new 60Hz videos. Based on previous tests with my 1024X location, doubling the frame count within each movie only increases the file size by ~30% using the identical Constant Ratefactor setting with the H264 codec. It appears Youtube also allocates a higher bitrate for 60Hz content, resulting in cleaner video with fewer artifacts compared to 30Hz. Also as 30Hz produces jerky movement with high speed zoom movies (necessary for extremely deep locations) I have found 60Hz zoom movies to be buttery smooth, with less motion blur effect if one follows a formation to the edge of the screen with their eyes.

[quaz0r]

yeah, cycling is based on the iteration values at each point, so each frame would need to be rendered based on this instead of just rescaling the same keyframe render.  and yeah, 60 fps is soooo much better than 30! 

[stardust4ever]

and yeah, 60 fps is soooo much better than 30!  

Agreed. And doubly so for Youtube with it's over-compression. Quality is still not quite as high as I'd like, but overall bandwidth is significantly improved over 30Hz. But 60Hz can be a double-edged sword. I've read complaints online of people whose PCs weren't up to snuff. I'm even getting a slight stuttering while viewing my latest Chromosome 1024X
http://www.fractalforums.com/movies-showcase-%28rate-my-movie%29/chromosome-1024x-%28720p-60hz%29-1-96-e1196/
in Firefox at 60Hz on my machine, but my bulldozer CPU is currently plowing through another AVI render so that could be a factor. My next video upload, "Simply Awesome II," is a complete rework of my first deep zoom mandelbrot video, 14:25 of buttery smooth 1080p 60Hz goodness. It utilizes pallet waves as opposed to infinite waves, and the color cycling combined with slope calculations really make it shine. I am unable to watch 1080p content at native resolution since I'm using an old 1600x1200 Syncmaster monitor from 2005.
My 1080p widescreen ASUS is currently set up in my game room, but the 4x3 is overall a better aspect for doing work anyhow.

I re-built my render machine in late 2011 with Windows 7 and an 8-core FX AMD processor, and since the new Gigabyte motherboard did not contain embedded graphics card, I bought an off-brand video card with 96 "cells." It was the cheapest card at the time which offered re-configurable mass-parallel processing cores to future proof my rig and enable the option for GPU fractal rendering. Some higher end cards at the time contained up to 1536 cells but also pulled 300 watts! My overclocked CPU with oversized heat sink was already turning my exhaust fan into a mini hair dryer when maxed out, and I didn't feel the need to add to the overall wattage of my rig since I'm not a PC gamer.

Oddly my giant desktop PC is over four years old and I see little reason to upgrade it since with the industry shift to mobile tech pushing cooler and more efficient CPUs while overall performance has been reduced or left unchanged. Newest desktop CPUs are only marginally faster that what was available in late 2011, and with months becoming like hours with series approximation, I have little need to upgrade at this time.

[quaz0r]

yeah, realtime playback of 60 fps h264 is kind of a tall order unfortunately.

the primary improvements i look forward to in new generations of cpus going forward is more cores and larger vector units, both of which intel has been stingy as **** with lately.  hopefully amd can get its act together and put out some interesting cpus to promote some healthy competition once again.  though like you say this situation is also exacerbated by the big focus on mobile. 

[Kalles Fraktaler]

with this "keyframe" approach to producing animations, i guess he is just rescaling each rendered keyframe to produce each frame when there is no cycling or anything?  whereas with cycling each frame needs to be re-rendered since each frame is different, which i imagine he has "optimized" about as well as one can expect.

Yes, this is the reason of the huge speed difference
(sorry for missing these posts)

[stardust4ever]

yeah, realtime playback of 60 fps h264 is kind of a tall order unfortunately.

Yet Best Buy have 4k TVs and UHD BluRay players/discs are apparently a thing now. Somehow they made custom low power ASIC logic chips capable of decoding an incredible 200+ million pixels per second (3840 * 2160 * 24fps =199,065,600).

If someone came out with a software UHD BluRay software player for desktop PCs, I somehow doubt the majority of users would have CPUs beefy enough to decode all that video, or GPUs and monitors capable of displaying it.

I downloaded a 4k 60Hz fractal zoom video off youtube and played it back on my 1600x1200 desktop monitor in WMP Classic. It played decent enough, but there were very noticeable compression artifacts even at my reduced 1600 pixel desktop resolution. What's the point in 4k video if you ruin it by encoding with much lossier compression settings resulting in increased visible artifacts? I swear 4k is a frigging placebo unless you are standing six inches away from the screen. I would rather simply have 1080p60 video encoded at UHD bitrates so that compression artifacts are nearly undetectable.

[quaz0r]

right, it isnt currently as incredible as it sounds.  it isnt actually decoding the incredible amount of information it takes to display a proper 4k picture because current tech cant really handle it yet.  all the media is severely bitrate starved garbage which is how they "solve" this problem of how to process that much information...just dont process that much information.  but go ahead and display it at that resolution anyway and hope everyone is too oblivious to know or care otherwise.  just like youtube crap is about 10 levels worse than the resolution it decodes to.  4k on youtube is worth playing back at 360p on a good day.  this is the direction everything will continue to go...market stuff as being bigger and better but make it a horrible pile of shite because all the braindead rubes wont know any better.

h265 was supposed to help solve this enormous problem of how to get a remotely respectable amount of quality into the bit starved trash that reasonably affordable current technology can handle...like most things it is taking much longer than hoped/anticipated to become a mature technology itself.  in the mean time most if not all of this media will continue to be garbage.

[stardust4ever]

right, it isnt currently as incredible as it sounds.  it isnt actually decoding the incredible amount of information it takes to display a proper 4k picture because current tech cant really handle it yet.  all the media is severely bitrate starved garbage which is how they "solve" this problem of how to process that much information...just dont process that much information.  but go ahead and display it at that resolution anyway and hope everyone is too oblivious to know or care otherwise.  just like youtube crap is about 10 levels worse than the resolution it decodes to.  4k on youtube is worth playing back at 360p on a good day.  this is the direction everything will continue to go...market stuff as being bigger and better but make it a horrible pile of shite because all the braindead rubes wont know any better.

h265 was supposed to help solve this enormous problem of how to get a remotely respectable amount of quality into the bit starved trash that reasonably affordable current technology can handle...like most things it is taking much longer than hoped/anticipated to become a mature technology itself.  in the mean time most if not all of this media will continue to be garbage.

Where's a "like" button when I need it?

Compression issues first became apparent to me with the DTV transition from analog between 2006-2009. I walked the isles at department stores and looked at all the LCD HDTV sets. Many store displays were still running analog cable content on the hidef sets, and the ones that got digital feeds to the TVs had compression artifacts.

My mom and I resisted subscribing to cable for as long as we could, but we live 2 miles south of the airport and Murphy's Law dictated the placement of the air traffic control tower directly blocking the path of the DTV over the air signals. Analog signal would create ghosting issues whenever planes took off and landed, resulting in a fuzzy but watchable signal. But DTV is extremely sensitive to multipath distortion, the picture would go black for 10-15 seconds, making OTA signals unwatchable, even with a properly aimed high gain antenna. It didn't matter that the signal strength within city limits was strong, when multipath distortion scrambled the signal resulting in all-or-nothing reception.

OTA DTV broadcasts cram up to 21Mbits of bandwidth into a 6Mhz band by transmitting a 3-bit signal with 8 individual logic levels at high bitrate. As a result, multipath destroys the signal. The "bonus content" sub-channels are almost exclusively broadcast in 480i, no better than analog cable, but worse with compression.

DVD is good.
BluRay is great.
4k UHD is placebo.

Here's why:
DVD has a maximum bitrate of 9.8Mp/s (720x480i 60Hz or 720x480p 24Hz NTSC, or 720x576 i50/p25 PAL) MPEG2.
BluRay has a maximum bitrate of 40Mb/s (up to 1280x720p 50/60, 1920x1080p24, or 1920x1080 i50/60) AVC.

The vast majority of BluRay movie content is 1080p24 progressive. Compared to (properly mastered) 720x480p24 NTSC widescreen DVDs, that is exactly 6x the pixel count per movie frame for BluRay compared to DVD. BluRay has a maximum bitrate of 40Mb/s compared to 9.8Mb/s for DVD, so content can be encoded at roughly 4x the bandwidth. Because AVC codec has roughly twice as efficient compression efficiency compared to MPEG2, this translates to roughly 8x video quality for 6x pixel count. This is good, because that means BluRay content can have a higher temporal quality afforded to it compared to DVD.

Additionally:
DVD dual layer capacity: 8.5Gb
BluRay dual layer capacity: 50Gb

This offers BluRay nearly 6x the storage space compared to DVD, so it is possible to store more content with less compression, even with regards to pixel count.

Moving onto UHD BluRay:

Screen Resolution: 3840x2160p, typically 24fps for movies, H.265, at 128Mb/s max bitrate.
100Gb triple layer discs; 66Gb dual layer.

First and foremost, I don't care what anybody says; 4k resolution is a placebo, and is slightly more data than a 20/20 pair of eyes can resolve at an optimal 60 degree view angle. And yes, if you look hard enough, you can still see the individual pixels from six inches away from the screen. Inversely curved screens are also a stupid gimmick. Great if you're sitting in the one seat in the home theater that forms an equilateral triangle with the screen edges. Move even a little bit to the side, and the curved screen distorts the view. I am extremely sensitive to shapes, and the curvature of the screen causing straight lines to become bent at non-optimal viewing angles, would drive me nuts. Flat panels are relatively viewable at all angles. Secondly, OLED tech is immature at this point, and current sets will suffer from burn in and will be junk in a few years. HDTVs are an investment. My first flat panel from 2006 is still kicking...

Statistically, you have a codec that is only marginally more efficient compared to AVS, compared to the quantum leap of MPEG2. The 128Mb/s bitrate allows for just over 3x the quality of BluRay, with 4x the pixel count. This is good. However, UHD-BDs can only hold 2x the data, so expect higher compression rates, especially for releases packed with hours of bonus content. Some newer films like the Hobbit Trilogy and others were filmed at 48Hz, while UHD BluRay does allow for higher frame rates than 24p or 30i. Again, the quality will suffer for long movies at higher frame rates crammed onto discs with only double the capacity. The real benefit will be from the higher bit depths compared to the 8bits per channel on older DVD and BluRay tech. This will allow a richer color gamut with blacker blacks and brighter whites compared to standard HDTVs. But you'll need a TV with newer display technology as LCD will not cut the muster. OLED is currently being pushed by manufacturers, which as I mentioned before, can have serious burn in issues within just a few weeks or months of constant use.

But most movies from the mid 90s until the end of the first decade of the new millennium, while encoded in HD, for many movies, special effects were not mastered in UHD or higher and did not take advantage of the superior bit depth either. As a result, conversions of recent movies, some even just a few years old, will be placebo. Movies originally mastered in HD will be upscaled to 4k but you cannot add detail where none existed. However just like with BluRay, old film movies contained much higher definition and dynamic range compared to SD, and benefitted immensely from being remastered to HD for BluRay reissue. A similar effect will occur for old classic film movies, with old photographic film negatives having a color depth and resolution far beyond even that of BluRay.

A bigger issue with the transition to 4k UHD is streaming and broadcast media. Modern RG-6 Coax used by cable companies has a total bandwidth of 2Ghz (0-2000Mhz) which has to be subdivided into hundreds or thousands of HD channels, as well as a significant portion of this bandwidth needs to be used for broadband internet access shared by hundreds of customers, at least for the last mile or so of access until said bandwidth is converted to fiber upstream.

Satellite is even worse. There are not only hundreds or thousands of HD channels beamed down from the geostationary satellite, but also every single local OTA station in the United States needs to be transmitted as well, because each customer needs access to local channels. As a result, most local channels available over satellite are heavily compressed 480i. I have seen compression artifacts in broadcast media, and at times said artifacts have even been plainly visible even using digital-analog RF adapters with old school CRTs! Ever since Comcast stopped transmitting analog channels over cable, overall quality has improved with less artifacts, largely due to freeing up a lot of available bandwidth.

I fear similar issues as cable and satellite systems upgrade their equipment to broadcast UHD content, the compression artifacts will get worse, not better. While disc based media obviously trumps broadcast media (also the variable bitrate on prerecorded media can be increased to peak rates much higher than average when needed, something that is absolutely not possible with constant bitrate required by broadcast or streaming), squeezing UHD content into networks already overloaded with current HD content, the artifacts will only get worse. If you allocate the same bandwidth to a UHD stream as an existing HD stream, the picture will look worse, sometimes much so. Artifacting in broadcast or streamed UHD content will become plainly visible even when downsampled to sub-HD resolutions.

Eventually bandwidth capability of equipment will catch up to UHD, but much more slowly. Moore's law is breaking down rapidly, as hard drive capacity and CPU speed are beginning to reach the limits as to what the laws of physics will allow. There may even be a global "crash" affecting servers and content providers world-wide, as technology and existing infrastructures cannot keep up with the rising global demand for bandwidth, storage, and computational resources.

Forgive me for my extremely off-topic rant...