Forgot Password
Pentax Camera Forums Home
 

Reply
Show Printable Version Search this Thread
02-22-2015, 09:10 PM - 1 Like   #1
Site Supporter
Site Supporter
lmd91343's Avatar

Join Date: Jul 2009
Location: Los Angeles, CA USA
Photos: Gallery | Albums
Posts: 1,027
Super resolution in post process

I used the technique discussed on the Photon Collective, as pointed out by Interested_oberver, to create a Super Reolution image.
Using my K3 at f4.5, 28mm, 1/400 at ASA 100 with my DA*16-50, and Photoshop CS6 extended, using 21 stacked images, with median averaging. I did the amount of sharpening on the super res photo as recommended in the article. In the standard res photo, I used as much sharpening as possible without artifacts showing. As a side note I find that I do much less sharpening with my K3 than my K5, K20 or K10.

This is a 4X Super Res. It is over 300 mb in size. It obviously displays the increased detail and no moire. That increased detail can be seen in the downsized Super Res displayed here. It does have the detriments of softer details of things in motion such as slightly sway tree branches, ripples in water, and people walking. Did I mention over 300 mb in size!

Take a look at the water ripples in the normal and super res photos.

The first pic is a side by side detail. The second is the super res. The third is a standard res. It has the detail area circled in red







02-22-2015, 09:35 PM   #2
Pentaxian




Join Date: Nov 2008
Location: Palo Alto, CA
Photos: Gallery
Posts: 2,497
Awesome, second (super res) image is indeed looking nice - crispier than the third image.

Just wondering however that benefits don't look so proportional to the effort in taking 21 stacked image. What's your opinion? It would be interesting to see how a good sharpening technique on third image alone will stack against the second image.
02-22-2015, 10:26 PM   #3
Otis Memorial Pentaxian
stevebrot's Avatar

Join Date: Mar 2007
Location: Vancouver (USA)
Photos: Gallery | Albums
Posts: 42,007
QuoteOriginally posted by yusuf Quote
It would be interesting to see how a good sharpening technique on third image alone will stack against the second image.
My thoughts as well. There is a general increase in local contrast that is obvious, but I am not sure that I am actually seeing more detail and not simply the illusion of detail.


Steve
02-22-2015, 10:37 PM   #4
Site Supporter
Site Supporter
lmd91343's Avatar

Join Date: Jul 2009
Location: Los Angeles, CA USA
Photos: Gallery | Albums
Posts: 1,027
Original Poster
QuoteOriginally posted by yusuf Quote
Awesome, second (super res) image is indeed looking nice - crispier than the third image.

Just wondering however that benefits don't look so proportional to the effort in taking 21 stacked image. What's your opinion? It would be interesting to see how a good sharpening technique on third image alone will stack against the second image. .
I think that the increased crispness/sharpness is related to the 4X increase in resolution, not the 21 layers. The extra layers provide the raw material for the increased resolution. I don't know how many layers are needed to provide the data for that extra resolution. Twenty-one layers may be overkill or not enough.

Twenty-one layers with 4X resolution and other post-processing took over one hour of my time and computer processing time. That was for an equivalent 100 megapixel sensor.

I intend to try 9X (220mp equivalent) and 16X (almost 440mp equivalent) with differing layer counts. However that will be when I have lots of extra time. It may take a day of processing time for each image tested.

---------- Post added 02-22-15 at 09:42 PM ----------

QuoteOriginally posted by yusuf Quote
It would be interesting to see how a good sharpening technique on third image alone will stack against the second image.
I tried sharpening the standard res photo 20 different ways. This is the most sharpening I could do without producing over sharpening artifacts.

---------- Post added 02-22-15 at 10:00 PM ----------

QuoteOriginally posted by stevebrot Quote
My thoughts as well. There is a general increase in local contrast that is obvious, but I am not sure that I am actually seeing more detail and not simply the illusion of detail.


Steve
I spent over two hours pixel peeping, viewing highly enlarged images comparing 4X resolution and standard on static items. The increase seems real to me.

I used the median layer averaging photoshop function. I cannot see how that function could change the overall contrast. To my eye, I see sharper transitions from one image element to another. There are four times as many pixels, allowing finer detail. The pixels are not duplicates of their neighbors, but each has an RGB value that is different than the one next to it.


Last edited by lmd91343; 02-22-2015 at 10:46 PM.
02-22-2015, 11:10 PM   #5
Pentaxian




Join Date: Nov 2008
Location: Palo Alto, CA
Photos: Gallery
Posts: 2,497
That's interesting to know, maybe you can share your sharpened pictures with different techniques too.

There is a definite advantage in the approach you followed, people have been using stacking for quite some time for the noise reduction, subject removal, depth of field etc. All I was saying here was that if it's worth the effort. If you like it, it's a way to go.

Another thing which needs to be taken care is if the subject moves, say leaves on tree moved due to the wind, it might be even counter productive.

in any case, thanks for sharing.
02-23-2015, 04:10 AM   #6
Forum Member




Join Date: Jan 2015
Photos: Albums
Posts: 83
subscibe to see later i'm interesting
02-23-2015, 11:23 AM   #7
Site Supporter
Site Supporter
lmd91343's Avatar

Join Date: Jul 2009
Location: Los Angeles, CA USA
Photos: Gallery | Albums
Posts: 1,027
Original Poster
another detail

Here is a detail of the drum bridge hand rail, slightly left of and below center.

It is at 200% for the super res and 400% for the standard res. The previous detail of the arcade was 100% super res and 200% standard res.



02-25-2015, 09:00 PM - 1 Like   #8
Site Supporter
Site Supporter
lmd91343's Avatar

Join Date: Jul 2009
Location: Los Angeles, CA USA
Photos: Gallery | Albums
Posts: 1,027
Original Poster
final analysis

This is my final analysis of my first attempt of creating super resolution images using post processing. I tried increasing resolution four, nine, and 16 fold.

DESCRIPTION:
The process is comprised of two parts:

The first consists of capturing many handheld frames, holding the camera as still as possible, This to provide the raw data to increase the resolution. The very slight unintentional movement of your hands and camera providing that data. The frames all have the same f-stop and shutter speed. The collection of the images takes a few seconds of high frame rate capture after setting the camera to manual.

The second is a six step post process that loads the multiple images as layers, resizes the image size to accommodate the increase in pixel count, align the layers, average the layers, flatten the layers, and sharpen the image. This is followed by all other processing such as sharpening or cropping.

LIMITATIONS
I found that I could not hold the camera still long enough to capture more than 21 near identical images. That is based on three image gathering attempts. I do have a slight tremor problem. However on my test day it was at a minimum. Given my best day before the tremors, I would not have been able hold still for 30 images. Processing time (mine and computer) is about an hour to a 4X image, about seven for a 9X image, and overnight for a 16X image. My computer is an Intel I7 with 16 gigabytes of memory and relatively high speed disks. The scratch disk is physically different the program and file storage disk.

PROS:
The 4X super resolution with 21 frames and layers, provided an excellent increase in sharpness far superior to standard, 9X, and 16X resolution. One hour of processing time is a small investment for the increased resolution. The 4X processing did not require paging of my computer's memory and intermediate file saves could be made at every step.

CONS:
A quick extrapolation for the number of the near identical images required for 9X and 16X resolutions would be almost 50 for 9X and 80 for 16X. I would find that impossible to achieve. The processing time would jump to near astronomical amounts. The 21 layered 9X and 16X attempts required paging. The 21 layered 9X and 16X intermediate step file save were impossible because the working file exceeded 2 gigabytes. Elements in the picture frame that moved during image collection were blurred or repeated.

RESULTS:
The 21 layer 4X super resolution image had resolution superior to the 21 layer 9X, 16X, and standard resolution. At first glance the 9X, 16X, and standard resolution appear to have the same sharpness at equal image sizes (16X 50%, standard res 200%, 9X 67%), however pixilation can only be seen on the standard resolution at those enlargements. I think that the 9X might be slightly better than the 16X and standard resolutions. All the super resolution images were created from the same 21 set of pictures/layers. The standard resolution is just one of those picture layers. In other words the same data was used for all.

I have presented a jpeg image of a detail of static and moving elements below. The details show equal image size. The 16X at 50% enlargement is first, normal at 200% enlargement is second, 4X at 100% enlargement is third, 9X at 67% enlargement is fourth. All are enlarged to the same size. There are two items to observe in each detail. The first is the bridge itself, particularly the post finials and the hand rail. The good trait, sharpness, can be judged from those. The bad trait, element movement, can be seen in examining the yellow/orange thin willow branch just below where it crosses the bridge deck. Look in the lower left hand portion of the image in black space beneath the deck. A single willow branch and splits into two at the lowest element. It is portrayed as a single branch in the standard res, as two in the 4X, five in the 9X, and two in the 16X.

CONCLUSION:
Given the difficulty in capturing more than 20 identical handheld images and the extreme process time for super resolution images greater than 4X, 4X super resolution is the greatest super resolution I would use. The image capture for 4X and one hour processing time are within my comfort zones. I imagine using this technique for hardscape landscapes on a regular basis.

IMAGE DETAIL:
16X at 50% enlargement, normal at 200% enlargement, 4X at 100% enlargement, 9X at 67% enlargement
03-01-2015, 10:24 AM   #9
Veteran Member
GeneV's Avatar

Join Date: Oct 2008
Location: Albuquerque NM
Photos: Albums
Posts: 9,830
I have been trying this technique as well, but for images with more noise. The largest number of exposures I have used was 9, and as few as 5 works well for noise reduction, but I am only going to 2x. I have some of the steps programmed as actions in PS so I can walk away, but I would expect 21 images to really bog the computer down.
03-01-2015, 12:53 PM   #10
Site Supporter
Site Supporter
lmd91343's Avatar

Join Date: Jul 2009
Location: Los Angeles, CA USA
Photos: Gallery | Albums
Posts: 1,027
Original Poster
QuoteOriginally posted by GeneV Quote
I have been trying this technique as well, but for images with more noise. The largest number of exposures I have used was 9, and as few as 5 works well for noise reduction, but I am only going to 2x. I have some of the steps programmed as actions in PS so I can walk away, but I would expect 21 images to really bog the computer down.
My 16 gig i7 takes an hour of human and computer time for 4x. What is your 2X processing time?

Do you increase image size by 40% in each dImension?
03-01-2015, 04:38 PM   #11
Veteran Member
GeneV's Avatar

Join Date: Oct 2008
Location: Albuquerque NM
Photos: Albums
Posts: 9,830
QuoteOriginally posted by lmd91343 Quote
My 16 gig i7 takes an hour of human and computer time for 4x. What is your 2X processing time?

Do you increase image size by 40% in each dImension?
Actually, I mistyped. I am multiplying the dimension 2x, so it is really a 4x increase in resolution.

I can get one done in less than 15 minutes at this point. I have 24 gig ram on an ASUS ROG. I now use ACR or Photoninja to convert RAW to JPEG first. I don't see a big difference in quality, and it makes the processing in PS much faster.
03-02-2015, 10:10 AM   #12
osv
Veteran Member




Join Date: Mar 2007
Location: So Cal
Photos: Gallery
Posts: 2,080
thanks for posting this technique.

need to see how it compares to shooting vertical panorama sequences that have been downrezzed.
Reply

Bookmarks
  • Submit Thread to Facebook Facebook
  • Submit Thread to Twitter Twitter
  • Submit Thread to Digg Digg
Tags - Make this thread easier to find by adding keywords to it!
camera, contrast, detail, equivalent, image, increase, k3, layers, mb, photo, photography, res, resolution, resolution in post, ripples, size, third, time, water
Thread Tools Search this Thread
Search this Thread:

Advanced Search


Similar Threads
Thread Thread Starter Forum Replies Last Post
Super resolution by pixel shifting in future Pentax cameras Mistral75 Pentax News and Rumors 190 03-29-2015 04:31 AM
Why post-process? Sigmund Troubleshooting and Beginner Help 135 05-16-2014 10:12 PM
Super resolution with K5 AlessandroB Pentax K-5 & K-5 II 6 04-26-2012 08:29 PM
Shake Reduction and super-resolution kadarpik Pentax K-5 & K-5 II 14 04-24-2012 02:27 AM
how to post process trevorgrout Digital Processing, Software, and Printing 6 08-10-2010 04:56 AM



All times are GMT -7. The time now is 11:19 AM. | See also: NikonForums.com, CanonForums.com part of our network of photo forums!
  • Red (Default)
  • Green
  • Gray
  • Dark
  • Dark Yellow
  • Dark Blue
  • Old Red
  • Old Green
  • Old Gray
  • Dial-Up Style
Hello! It's great to see you back on the forum! Have you considered joining the community?
register
Creating a FREE ACCOUNT takes under a minute, removes ads, and lets you post! [Dismiss]
Top