Average image, well kind'a

trandoductin · **Posted:** Wed Oct 07, 2015 2:34 pm **(#71)**

patdavid wrote:

Yep, wasn't the friends video just all of the episodes played over each other (audio included)?

yeah, i think that was it... did you watch it and did it give you a headache?

trandoductin · **Posted:** Wed Oct 07, 2015 2:44 pm **(#72)**

patdavid wrote:

Not sure if you folks had seen these... (All of my Amalgamations on Flickr).

They were from a few years ago. I used imagemagick convert initially to handle the blending (because I could do much larger image sets faster than trying to adjust opacity on a per-image basis in GIMP).

To verify the results against Salavons initial work years and years ago, here is my result of the Playboy centerfolds (All of the 1970's):

[ Image ]

Which I then began extending to other things, like all of the Sports Illustrated Swimsuit Edition covers:

[ Image ]

I then moved on to photographer Martin Schoeller's close up portraits:

[ Image ]

My two favorites were George Walken/Christopher Clooney:
[ Image ]

And all the "Dreamy" men in the list:
[ Image ]

From there I basically started looking at much longer averaging sets, and settled into music videos. I would dump all of the frames of the music video and use a custom script for G'MIC to mean average all of the images (well over 5000-6000 images per video):
http://blog.patdavid.net/2013/09/mean-a ... ideos.html

[ Image ]
[ Image ]
[ Image ]

A little while back I had a look at the average movie covers from Netflix that were suggested to me by their algorithms:
[ Image ]

Apparently, doing puny music videos wasn't quite enough for me. Here's where I'm currently focusing some efforts:
[ Image ]
[ Image ]
[ Image ]
[ Image ]
[ Image ]
[ Image ]

A couple of my original posts from back in 2012 for reference (if anyone wants to follow along):
http://blog.patdavid.net/2012/08/imagem ... files.html
http://blog.patdavid.net/2012/08/more-a ... eller.html

seeing these makes me itch to try ImageMagick to do it to a larger sets of images now

(damn you PatDavid! shake fist at sky)

Issabella · **Posted:** Wed Oct 07, 2015 3:19 pm **(#73)**

Tin, you will get very good outcomes in your ImageMagick tries

RobA · **Joined:** Jul 14, 2010 **Posts:** 697

A while back I came across this page of gimp plugins for astronomy:
http://hennigbuam.de/georg/gimp.html

Merge Layers might be useful here:

Quote:

Merges all layers. You can choose, which method for the mean value you want: Arithmetic, geometric, median, sigma median (1 or 2 pass).
Sigma median is a median of all values inside [median-sigma;median+sigma] with sigma the standard deviation.
The pixel values are weighted by alpha value (for arithmetic and geometric mean value), and ignored (for all median mean values) if alpha channel is 0.
You can mulitply each pixel value by a given constant and then add a given constant (before calculating the mean value!).
A preview widget lets you decide which is the best method for you.

Haven't tried it...ymmv

-Rob A>

trandoductin · **Posted:** Wed Oct 07, 2015 10:07 pm **(#75)**

RobA wrote:

A while back I came across this page of gimp plugins for astronomy:
http://hennigbuam.de/georg/gimp.html

Merge Layers might be useful here:

Quote:

Merges all layers. You can choose, which method for the mean value you want: Arithmetic, geometric, median, sigma median (1 or 2 pass).
Sigma median is a median of all values inside [median-sigma;median+sigma] with sigma the standard deviation.
The pixel values are weighted by alpha value (for arithmetic and geometric mean value), and ignored (for all median mean values) if alpha channel is 0.
You can mulitply each pixel value by a given constant and then add a given constant (before calculating the mean value!).
A preview widget lets you decide which is the best method for you.

Haven't tried it...ymmv

-Rob A>

Interesting script (that Merge all).
here are results for Arithmetic, geometric, median, sigma median (1 or 2 pass).
from 49 images of Jesus

I did bump up the contrast to make it pop more ..but i think i am leaning toward the sigma median ones.

patdavid · **Posted:** Thu Oct 08, 2015 10:59 am **(#76)**

These actually remind me of my personal favorite of all of my blends:

Mr. Presidents

Mr. Presidents by Pat David, on Flickr

A mean/median blend of all of the U.S. Presidential portraits. I'm actually going to print this really large soon and hang it in my house.

trandoductin · **Posted:** Thu Oct 08, 2015 11:56 am **(#77)**

patdavid wrote:

These actually remind me of my personal favorite of all of my blends:

Mr. Presidents
[ Image ]
Mr. Presidents by Pat David, on Flickr

A mean/median blend of all of the U.S. Presidential portraits. I'm actually going to print this really large soon and hang it in my house.

nice the colors remind me of old portraits like something from the harry potter movies
here's one from about 80 images of buddha.

trandoductin · **Posted:** Fri Oct 09, 2015 11:36 am **(#78)**

I wrote a very simple script to aid in scaling layer in order to align images.
First rotate your working layer so that it's straight, have two guides (left guide and right guide) for the left and the right eye of the reference image ,
Then move the layer you want to scale to fit so that left eye is lined up with left guide.
You then simply supply the script with x-coord of left-guide and right-guide, and your the new-right x-coord which is the x-coordinate of the right eye of the layer you want to scale so that it'll scale the layer to the right size so that you when you move and match up left eye to left guide, right eye will line up with right guide.

Attachment:

auto-scale.scm [1.66 KiB]
Downloaded 146 times

Script will be availabe under script-fu/Auto scale...

trandoductin · **Posted:** Fri Oct 09, 2015 8:48 pm **(#79)**

average of 115 images of Buddha.
used eyes to align (I used the above script) but it still involved a lot of tedious repetitive work.

2-ton · **Posted:** Fri Oct 09, 2015 10:37 pm **(#80)**

It is beautiful

trandoductin · **Posted:** Sat Oct 10, 2015 12:46 am **(#81)**

2-ton wrote:

It is beautiful

Thanks 2-ton

AnMal · **Posted:** Sat Oct 10, 2015 4:38 am **(#82)**

trandoductin: that one is lovely! i see you've posted a script too, looks very useful!

Wallace · **Posted:** Sat Oct 10, 2015 5:31 am **(#83)**

trandoductin wrote:

average of 115 images of Buddha.
used eyes to align (I used the above script) but it still involved a lot of tedious repetitive work.
[ Image ]

Nice Buddha image.
:bigthup

trandoductin · **Posted:** Sat Oct 10, 2015 10:15 am **(#84)**

Wallace wrote:

trandoductin wrote:

average of 115 images of Buddha.
used eyes to align (I used the above script) but it still involved a lot of tedious repetitive work.
[ Image ]

Nice Buddha image.
:bigthup

Thanks Wallace

trandoductin · **Posted:** Sat Oct 10, 2015 11:15 am **(#85)**

AnMal wrote:

trandoductin: that one is lovely! i see you've posted a script too, looks very useful!

yeah i wrote the script so that i don't have to switch program to spread sheet and manually enter percentage to scale...trying to be as lazy as i can

trandoductin · **Posted:** Sun Oct 11, 2015 2:37 pm **(#86)**

273 images of Eiffel tower (not aligned at all).

2-ton · **Posted:** Sun Oct 11, 2015 6:29 pm **(#87)**

I like the aligned average images, not so much the non-aligned ones.

trandoductin · **Posted:** Sun Oct 11, 2015 8:08 pm **(#88)**

2-ton wrote:

I like the aligned average images, not so much the non-aligned ones.

me too..
but non aligned images sometimes surprise me more, with aligned images i can sort of imagine what it's going to look like since features that are aligned with emerge much more than non aligned features.
here's another non aligned one (257 images) of "Brown hair".

RobA · **Joined:** Jul 14, 2010 **Posts:** 697

trandoductin wrote:

here's another non aligned one (257 images) of "Brown hair".

Are you resizing these all the same before averaging them?

I was wondering if this could make a poster. Doing at this size then scaling the result would be faster but would have enlargement artefacts. Would scaling them all up first then running the average eliminate such artefacts?

-Rob A>

trandoductin · **Posted:** Mon Oct 12, 2015 2:43 pm **(#90)**

RobA wrote:

trandoductin wrote:

here's another non aligned one (257 images) of "Brown hair".

Are you resizing these all the same before averaging them?

I was wondering if this could make a poster. Doing at this size then scaling the result would be faster but would have enlargement artefacts. Would scaling them all up first then running the average eliminate such artefacts?

-Rob A>

I resize them all to 1000x1618 or 1618x1000 then average them.

	Similar Topics	Replies
	Spin-art Grid kind'a like Kandinsky's circles but for any image	1
	Get average color of selection	6
	How to acheive this kind of effect (the holographic effect)	5
	Retexture an image with another image as "Image Pattern" Plug-in	4
	Show image on web browser to fill 100% width AND image map	2