https://www.youtube.com/watch?v=-hRBzBnzkX0

 


 

The lion kingdom set to the task of stabilizing the extremely shaky crew-11 re-entry footage.  It was not possible to stabilize with Lucas Kanade.

After decades of manually motion tracking flying objects, the lion kingdom pondered yet another take in the motion/tracking/stabilization department.  Manely, a hack to just find the left most blob in the image which meets a minimum size & move its leftmost point to a fixed position.  If nothing qualifies, it copies the offset of the previous image.

Another idea was to scan every column from the left, find all the pixels which exceed a certain threshold, track the center of the 1st hit when a column has enough qualifying pixels.  Lions believe this would have too many false positives from noise & a blob size which factored in horizontal information would be required.  It might  be simpler to use the column algorithm though.

A variation was to find the biggest blob. 

 The user would keyframe the destination position.  It could use the output of the chroma key, threshold plugins to define the hit blob.

The minimum size blob failed from false positives & the largest blob failed when the fireball tail was bigger than the fireball & a separate blob.   

The next idea was to track the brightest window.  That was highly dependent on under exposing.  It worked when the fireball was viewed from in front of the tail & failed when it was viewed from behind the tail.  The brightest window algorithm rarely hit an area bigger than 1 window.  It suffered from cases where a larger part of the window was filled by dimmer pixels in the tail & a smaller area of brighter pixels filled the window in the fireball.  It also suffered from the inability to track a single leftmost pixel.

 

Largest blob + brightest window managed to overlap enough to stabilize the whole video for the 1st time.  The result was most impressive.  Would try 4x the shutter speed & wide open to reduce the motion blur & get a hair more highlights.  Would continue using movie crop mode to get the unevenness of the tail.  The existing movie was ISO 6400, 1/30s, F4.0, movie crop mode, 200mm, 1.6x sensor, AWB.  F 2.8 would double the brightness.

Darker exposure would risk not having a big enough blob to track, though.  Probing with the eye dropper showed the tail to quickly drop to .5 while only the fireball was maxed out.  During peak heating, there's wasn't any more detail to be had in the tail, though the fireball might have more flashing.

 Still photos would capture more of the tail with the same detail, but wouldn't convey its speed.

---------------------------------------------------------------------------------------------------- 

Another algorithm could take the brightest average pixel in a window or blob with a minimum size of the blob instead of the largest blob.  Brightest blob would use the chromakey track for blob detection & the target track for brightness detection.

Helas, the 1st problem is blobs contain a lot more dark pixels than fixed windows, so the dark pixels end up lowering the average.  It has to use the brightest pixel in the blob.

This would suffer from the same false positives as the left most blob over minimum size again but it would have the advantages of the brightest window algorithm.  All these algorithms continue to track the leftmost pixel because that has been more stable than tracking a center pixel.


Largest blob, brightest blob, & brightest window compared.  It now uses blob layer to define the blobs & target layer to detect brightness & apply motion.

 The brightest blob algorithm managed to do the entire video itself. Instead of supporting brightest window, it should sum the n brightest pixels in the brightest blob to get the brightness value.  It would be different from the minimum blob size.  For now, the single brightest pixel seems to be enough complexity.

 


 Still neglected are the side it tracks & the destination coordinate.

 Making the threshold more selective made it a hair more stable.  Helas, no amount of audio EQ could detect any sonic boom.  Neither were there any recordings of a sonic boom near the coast.

 ---------------------------------------------------------------------------------------------------------

After finally disposing of the lion kingdom's beloved 2nd PC case & spotting just how much Manley Man spends to store physical vinyl, lions began pondering finally beginning to dispose of some of the physical media.  The prime candidates are audio tapes which are now on gootube in higher quality.  Helas, most of the media is home recordings.  Could a lion live without the odd ways certain songs were recorded from the radio, the DJ overlaps, interjected sound bytes?  Young lion added odd sound bytes to make use of every last bit of tape.  The later ones were straight recordings from CD's with nothing else interjected.

Another problem is blank audio tapes are quite valuable to other animals, since they're no longer made.  There's a very very very small chance of lions using them to store tiny amounts of backup data.

 That 2nd PC case was the beginning of life without real PC cases.  It wasn't big enough for the cooling it needed, so the power supply always went on the outside.  In its later years, it was a bathroom table.  After toilet duties went to a phone, it was relegated outside the bathroom.  Disposing of it renewed the feeling of nearing the end & the coming big move.

 The other things rising up the list are the DVD towers.  With lions planning on renting forever, they have little chance of ever being assembled again.  The loading tower has already begun shedding parts & is the 1st candidate.

 ------------------------------------------------------------------------------------------------------

 

 https://www.instagram.com/marguerrotype/


 Need to be restored by that.  A hippie hottie would have gotten along with lions, if only they didn't need breadwinners.

https://www.youtube.com/watch?v=9W6msaAQiIk

 


Got lions pondering algorithms which generate music from photos of hottie trophy wives.  Maybe they should follow rules of harmonics at least & convert the photos to frequency domane.  Years ago, wavelet based music generation would be the answer.  Nowadays, they would use the output of a conv net or use the image data as a seed value in a generative algorithm.  Lions still doubt any image data could be turned into useful music.

 

https://www.youtube.com/watch?v=5ohakQ9DE-c 

 

 Speaking of racks, mexican canadians are 1 of a kind.

 

 

 

 

 

 

 

 

 

Comments

Post a Comment

Popular posts from this blog

snow white

Image
    It was time to take on the lion kingdom's 1st 1 star movie. It honestly wasn't as bad as the Dictator, Bedazzled, Monty Python.  Most of the hate was from comparing it to the classic.  Visuals were good for animals who are into castles.  Wonder Heroine as the villain was a good twist.  It was bat shit boring.  Tedious jobs like dialog writing are all automated now & that's definitely not as bad as other movies.  Automated 3D model generation has made modern visuals very lavish.  If only that level of detail could be done in realtime for VR goggles.    Zegler was back from her previous engagement in district 12.   The mane redeeming factor for lions was evil Wonder Heroine.     & finally the Joan of Arc scene everyone hated.    & the ending just fizzled.  It seems they're trying to depart from the physicality of past heroine warriors & make them more like housewives.  It's h...
Read more
Image
   Amusing that Grok applied an IPv6 style pseudo header to the computation of an IPv4 ping packet, even though IPv4 doesn't use the pseudo header.       Another Grok error was in a basic TCP demonstrator.  To avoid a TIMED_WAIT on the server blocking the port, you have to exit the server without closing the socket.   These errors take a lot more time to track down than doing it the old way.  It might be fixed in paid versions of Grok. The solution to everything ------------------------------------------------------------------------------------------------------------------- The 2nd Krispy Kreme run.  Always a satisfied mane after 6 miles.  Always a horrible bloating, after 13 miles.  Noted the half marathon times with mane feedings are on pace for the lion kingdom's best full marathon time.  Lions took a step down from their maximum power delivery in the shortest time when the leash enabled food breaks.  Before then...
Read more
Image
 Finally tracked down the Sun bootloader font from many decades ago & put it in Cinelerra as gallant12x22.  Put Sun Gallant Demi into the goog to get useful results. https://github.com/dim13/gallant/tree/master/ttf   It's a fixed point font in an archaic format.  Some guy converted it into a TTF with a lot of antialiasing.  He has a PNG image showing no antialiasing.  It has to be set to size 22 & requires a threshold to get rid of the antialiasing.  Freetype might be shifting it 1/2 pixel vertically. It had a kind of curiosity value, but lions have had no need for any fixed point fonts besides the 24pt Sony of CAEDM fame. Another hit said it was 1 of the fonts available in kernel framebuffer mode.  Lions haven't used kernel framebuffer mode in decades.  It could possibly replace 24pt Sony if lions wrote a fixed point converter. As for the other Sun lore, there's some potato cam footage of the Sun flight simulator, Aviator.  You ...
Read more