January 30, 2015, 04:47:28 PM

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Neutral

Pages: [1] 2 3 ... 8
1
Already there....

Autopano Giga (Image stitching software) uses GPU cores to accelerate image processing. It makes a HUGE increase in speed...

Autopano Giga  ( http://www.kolor.com/image-stitching-software-autopano-giga-photo-stitching-feature.html ) seems to be very attractive product - seems to be much superiors compared to stitching in Photoshop

Are you using it and if so what is your feedback on that ?
It is not cheap but seems that product cost is for real  value


2
...when it is possible to utilize more than 1500 processors on latest NVIDA cards for processing instead of just 4 or 8 cores on main CPU.

Just to level expectations - just because modern GPUs have 1500+ cores doesn't mean that you'll gain a 375x increase in performance over a quad core machine by utilising them.
<...>

  Sure, I was not telling that using additional 1500 CUDA cores compared to 4 main CPU cores would give 375x increase in performance.  Single CUDA core and single main CPU cores have different processing power and available to them resources and their main target applications are different.
 
   What I meant that using full CUDA processing resources using all available CUDA cores could provide drastic improvement in performance for RAW image processing.  Especially could be useful for image denoising  which could be splitted into huge number of parallel processes  using separate process for each small image block.

   If on my laptop PRIME DXO denoising  takes  about 150 seconds  for A7R,  80 seconds  for 1DX and ~23 seconds for a7s RAW files  and I see  this process is using all  4 main CPU  cores (8 threads) on laptop up to 100% and  boosts  CPU clocking up to 3.4Ghz and CPU temperature jumps to 100C  then using  1536 cores of NVIDIA GTX 780M could provide drastic performance improvement and reduce load on main CPU.  Even 10X better performance would result in 15sec for a7R, 8 sec for 1Dx and 2 sec for a7S with prime denoising and I believe that could be even much better fully utilizing all CUDA cores

  When I started this topic I provided several inks to the very interesting presentation regarding the subject, one is NDIVA presentation and the other one of the very impressive real time embedded image processing implementations.  That  papers shows what level of performance   improvement   could be achieved using CUDA technology for image and video processing directly from RAW file .

Benchmark results for Fastvideo  industrial cameras implementation with  real time processing are really amazing:
http://www.fastcompression.com
http://on-demand.gputechconf.com/gtc/2014/presentations/S4728-gpu-image-processing-camera-apps.pdf
====== 
Final Benchmark on GPU (Titan)
CMOSIS image sensor CMV20000, 5120x3840 (~20mpx), 12-bit, 30 fps
   GeForce GTX Titan GPU
   Host to device transfer ~1.5 ms
   Demosaic ~3.1 ms
   JPEG encoding (90%, 4:2:0) ~7.8 ms
   Device to Host transfer ~1.3 ms
   Total: ~13.7 ms
P.S. This is the benchmark for PCIE camera CB-200 from XIMEA
-------
Solution for Photo Hosting
   Task description: load-decode-resize-encode-store
    Image load ~1.5 ms for 2048x2048 jpg image
   JPEG decoding ~3.4 ms
    Downsize to 1024x1024 with bicubic algorithm ~0.7 ms
   JPEG encoding (quality 90%, 4:4:4) ~3.4 ms
    Image store ~1.0 ms
    GPU processing time ~7.5 ms
Total time ~10 ms
--------------------

I wish to see that level of performance in products that I currently use, especially in some new LR release.

And more info in NVIDAI presentation:
http://on-demand.gputechconf.com/siggraph/2013/presentation/SG3108-GPU-Programming-Video-Image-Processing.pdf


3
I am wondering when image processing softwares (LR, DXO, C1 etc.)  will start using GPU for image processing .
For C1 that would be quite a while ago with continuing improvemts over the recent years/releases.

Yes , you are right, C1 and DXO uses that to some extent (with GPU acceleration enabled) but I am not sure how efficiently they utilize full GPU power. My understanding that mostly for image rendering and less for RAW processing.
But actually both are pretty fast on my laptop with NVIDIA GTX 780m card

Here are some interesting tests results for C1:
http://diglloyd.com/blog/2014/20140117_1-CaptureOnePro-GPU.html

But LR is still not using GPU acceleration and not sure if that this  will be utilized in LR6.

4
Interesting how long we will be waiting until Adobe will start using GPU acceleration in LR  (including RAW processing not only Open GL for smother  rendering ) to improve performance especially NVIDIA CUDA processing

This is possible and there are some implementations providing extremely fast processing: 

http://www.canonrumors.com/forum/index.php?topic=24728.0


5
 I am wondering when image processing softwares (LR, DXO, C1 etc.)  will start using GPU for image processing .
 This could drastically increase performance and processing capabilities and could help to implement more complicated and more resource demanding algorithms. Especially using NVIDIA CUDA – when it is possible to utilize more than 1500 processors on latest NVIDA cards for processing instead of just 4 or 8 cores on main CPU.
One who first implement this could have great advantage over other competitors.

This question comes to my mind each time when there are news about new major S/W releases – e.g. now with the information that Lightroom 6 will be coming soon.

Earlier Adobe was telling about difficulties to implement parallel processing but this does not reflect current realities.

Simple search on WEB shows that there are patents existing for GPU image processing as well as number implementations and API Libraries to utilize CUDA technology for image and video processing including RAW files processing  and some implementations which provide amazing processing speed.

Here are some references:

1. GPU  Raw image processing patent  US 8098964 B2
http://www.google.ca/patents/US8098964

2. http://www.ximea.com/de/technology-news/gpu

3. http://on-demand.gputechconf.com/siggraph/2013/presentation/SG3108-GPU-Programming-Video-Image-Processing.pdf
------
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=5&cad=rja&uact=8&ved=0CDUQFjAE&url=http%3A%2F%2Fon-demand.gputechconf.com%2Fsiggraph%2F2013%2Fpresentation%2FSG3108-GPU-Programming-Video-Image-Processing.pdf&ei=cfjBVOPfFpDiav_0gcAF&usg=AFQjCNEZ78COMnMT4hvBZrflwN3-b_ZibQ&bvm=bv.84349003,d.d2s
------
http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=5&cad=rja&uact=8&ved=0CDUQFjAE&url=http%3A%2F%2Fon-demand.gputechconf.com%2Fsiggraph%2F2013%2Fpresentation%2FSG3108-GPU-Programming-Video-Image-Processing.pdf&ei=cfjBVOPfFpDiav_0gcAF&usg=AFQjCNEZ78COMnMT4hvBZrflwN3-b_ZibQ&bvm=bv.84349003,d.d2s


6
I would suggest to have a look at this :
   http://www.originpc.com/workstation/laptops/eon15-s-pro/
    http://www.originpc.com/workstation/laptops/eon17-s-pro/

ORIGIN PC is US based PC manufacturer/vendor for professional top level grade laptops and desktops which are fully customizable - could be ordered in any required for customer configuration.
You just need to select laptop , select required configuration, place order and you will get best for your budget.

I ordered  one in my own configuration one year ago and I am extremely pleased with that since I received that.

I spent quite amount of time evaluating different options  from different vendors and found that ORIGIN PC is the best way to go - best investment for available money - you get exactly  what you want and need .
It is not very well known brand - as they specialize not in mass production but in providing custom made system with top performance

Regards.

7
Software & Accessories / Re: Photo Editing Laptop Recommendations
« on: January 13, 2015, 12:43:28 PM »
I would suggest to have a look at this :
   http://www.originpc.com/workstation/laptops/eon15-s-pro/
    http://www.originpc.com/workstation/laptops/eon17-s-pro/

ORIGIN PC is US based PC manufacturer/vendor for professional top level grade laptops and desktops which are fully customizable - could be ordered in any required for customer configuration.
You just need to select laptop , select required configuration, place order and you will get best for your budget.

I ordered  one in my own configuration one year ago and I am extremely pleased with that since I received that.

I spent quite amount of time evaluating different options  from different vendors and found that ORIGIN PC is the best way to go - best investment for available money - you get exactly  what you want and need .
It is not very well known brand - as they specialize not in mass production but in providing custom made system with top performance

Regards.

8
Here is simple example when sensor DR and sensor  IQ are very useful.

If you think the second image is an improvement then we are talking about different things.

+2, the second image is why HDR also stands for Horribly Divergent from Reality.

1.   Interesting enough that all who replied did not get to the main point of the post and immediately started to tell how ugly the second image is))).  But second one was not main point of the message but just previews of how initial image  was pushed in processing and preview was done by windows snipping tool and was not supposed to illustrate anything – just to give idea to which extent initial image was pushed in processing.
     Main point of the post was different and was illustrated by 3 image and that was that when sensor has good DR and IQ it would allow to get similar results as multi-shot HDR images without any efforts and time spent and allowing to push hard processing limits without introducing any visual processing artifacts like noise, halos etc.  even when image is “overcooked”.

  So this was illustrated by 3d image (100% crop of the second image) showing that even when image processed to the limits it does not have noise, halos on extreme contrast transitions and it resolves details up to the every pixel.
 
2.  Example itself was just “Technical example” using image which mostly contains small details with uniform colors across the frame  not clearly seen on small image especially on preview done by windows snipping tool. 
This is kind of images are most difficult for HDR and could be seen well only with high enough resolution. 
Best for HDR are shots with big enough objects of different colors  in the frame.
So now I attach better preview of the second image in the first post   - directly exported with better resolution.
The same is for 100% crop which is the illustration of the main point of the post explained above- now directly exported from image processor
Also one more image that was not pushed so far and processed to have more realistic look – just compressed DR to fit visual limits – no black or white clipping and a little bit of contrast added in shadows.
Probably most will find that more pleasant for their eyes.
 And also 100% crop of it – to illustrate once more that was explained above about absence of visual processing artifacts usually found on muli-shot HDR images 
Click on any image to open it and see with better resolution.

  In any case all that images should not be considered as final images - that is just first step of processing   -  DR  compression to fit visual range.   After that person could adjust it  to personal taste by doing selective color editing (contrast/saturation/luminance) differently on highlights, midtones and shadows and doing other kinds of tone mapping  processing using most suited tools for that.

Hope all that would clarify what I was trying to say in my previous post

9
Here is simple example when sensor DR and sensor  IQ are very useful.
Just simple shot in autumn forest.

This is single shot exposed for the sky.
Done by using A7R with Canon TSE-17

First image – as shot, no adjustments in LR5 snapshot on LR5 screen (full image).

Second one is one processed in DXO PRO 10 – done without any efforts just in few seconds to do few adjustments. This snapshot of the processed file as seen on the windows image viewer
 For me DXO Pro 10 it is much better tool than LR5 for such kind of images – it is has separate adjustments for micro contrast, fine contrast (in LR only clarity)  and it also allows to adjust contrast separately in highlights, midtones and shadows. In LR5 it is much more difficult  to get result close to this

Third one is 100% crop from the most contract image area to see image IQ and quality of shadows recovery.
Image DR and IQ matters much for me as it save my time and efforts in getting quickly required result.

For low light all around camera my current favorite is A7S - it make possible to do things that were almost impossible before.
For events/sport and action 1Dx is still the best. All cameras just complement each other in different situations.
I have a feeling that many 1DX owners following the same route. 

10
 As Jrista mentioned in his previous post   it would be interesting to see comparison of normalized to 8mb images from NX1, 7DM2, 5Dm3 and A7S.  Attached are ones at ISO6400, ISO12800 and ISO 25600.
  Normalized images comparison just illustrates what was said before about NX1 noise characteristics  which are better than 7Dm2  and  on normalized images comparison  it is even more obvious that NX1 is much less noisy at high ISO than 7Dm2 and not very  far behind of 5Dm3.  To my opinion for general photography NX1 is much more advanced camera and has more potential than 7Dm2 and I definitely I would recommend NX1 in favor of 7Dm2 if someone asks my advice.  Also NX1 AF tracking capabilities looks much more advanced to me than 7Dm2 AF system adopted from 1DX.   NX1 AF tracking system looks like something borrowed from military object tracking systems used in optical weapon object recognition and targeting system on missiles and military jet fighters. Maybe Samsung does something this area and they used that experience this in NX1
    I have feeling that now when semiconductors giants like Samsung and Sony are on the market with the latest imaging sensor technologies and their resources in R&D and manufacturing capabilities and the rapidly increasing speed of new technologies development and implementation Canon just cannot compete any more on the sensor market (which is also semiconductor technology).  Even if Canon would develop something in their R&D LAB they still need to manufacture that and this is where is their big weakness (manufacturing processes/technologies etc.) which is very difficult to overcome and they cannot run long in semiconductors technology race (as this is not their primary business), unless they ask some other big player (e.g. Samsung) to manufacture their chips, like Apple was doing for their iPhones and iPads.
   I would be interesting to see (just my wish) Samsung coming on MF sensor market and possibly build up their own affordable ML MF camera.  They just can buy one of the MF makers (e.g. Hasselblad) and enter market easily with top-notch sensor technologies.
  On comparisons shots is also could be seen that 1Dx and A7S are almost on par with each other, maybe A7S a bit better.   But practically (from my experience) A7S images are easier to clean and they have more margin for post processing (could be pushed more) before exhibiting any problems.
  Also there are couple of comparisons with P645Z.
   Images from P645Z almost looks one stop better that 1Dx or A7S at ISO 12800 and ISO25600.
So far nothing can compete with latest MF sensor from Sony for low light performance.
Hopefully we will see soon ML MF camera from Sony.

11
That's my point. Statistically, the NX1 has lower noise, which leads to richer color and contrast (hence the reason the parts of the NX1 image that are supposed to be dark look dark! :P)

It does have very slightly lower noise, but that is not the reason it's a darker image which is what leads to the difference in our perception of color and contrast.

Again I'll note that with color NR the 7D II ends up looking a tiny bit cleaner, i.e. lower luminance noise. But in the end neither requires a different work flow or more work. Neither has "better data."

My perception is a bit different and I tend to agree with  Jrista.
   My comparison of RAW samples (no NR at all - no Chroma no Luminance  NR)  at DP comparison tool   which I posted a bit earlier shows that at  ISO3200 NX1 RAW samples visually  look better/cleaner/crisper than ones from 7Dm2  and as I also noted NX1 noise pattern is visually better and more pleasant than noise pattern of 7Dm2. NX1 is less blotchy , has a bit more higher noise frequencies and a bit more regular  - so should be easier to clean up compared to 7Dm2 . Blotchy noise pattern from many Canon Cameras is something that was always irritating  me as well as low performance at low contrast details in red channel - just smearing them away. 1Dx fortunately is better in this respect though also suffering a bit.   
Also on this samples NX1 is not downsampled/normalized  to 7Dm2 resolution which would also add more difference in favor of NX1.

  I was interested to see if NX1 could compete with A7S - but so far A7S is far ahead of all the competitors.
Just see here  as an example one of  test picture done handheld by A7S in extremely low  light  conditions at ISO20000.
One small screen snapshot and also  full image exported from LR



12
When I read Lula review of Samsung NX1 sensor I was really impressed and I was much interested to see how it compares with other cameras (including FF). I am also as much impressed as MR and Jrista about NX1 sensor technology, not only low noise level but by AF capabilities and the fact that AF points cover 95 percent of the screen.
   This AF tracking technology is my dream for many years for sport/action medium format camera – especially for acrobatic events. With Canon 1Dx and tele zoom lens I can get close to the performer but movements are so fast and so erratic that it is not possible to track that manually – performer always jump out of the view. With MF camera (around 80mp) and NX1 tracking technology it is possible to have much wider angle of view , camera will track object without need to move camera and then you can do required crop  of the shot and get high quality resulting image .  With this it would not be required to use long tele lens. I hope to see that in a couple of years in coming Mirrorless MF cameras
Now back to NX1 sensor noise levels at high ISOs and comparisons with Canon 5Dm3, 7Dm2 and Sony A7.
I did that using DP Review studio –shot comparison tool and below are screen snapshots for comparisons of Samsung NX1, Canon 7DII, Canon EOS 5D M3 and Sony A7S, and also one with Sony A7 at ISO3200 from.
-   At  ISO3200 NX1 is very close to 5Dm3, a bit more of luminance noise but noise structure is better than 5Dm3 – less blotchy and easier to clean up.
-   At ISO6400 5Dm3 looks better than NX1 which could be expected for FF compared to APS-C sensor.
-   At both iso3200 and ISO6400 NX1 is better that Canon 7DM2
-   King of low light in FF segment is still Sony A7S, at ISO6400 it has less noise than Canon 5Dm3 at ISO 3200..
At ISO 12800 A7S is also noticeably better than 5Dm3 at IS06400.
From comparison it is easy to see that A7S is about 1.5 stop better than Canon 5Dm3.
   I bought A7S recently (could not resist temptation) and since then enjoy it to great extent – it makes impossible possible especially in combination with DXO Optic Pro 10 PRIME noise reduction. Getting very clean images (in shadows areas) that were shot in very dim light at ISO up to 20000 which was almost unbelievable even 1 year back. Now using 1Dx much less than before.

 What is interesting and bit surprising that at ISO3200 and especially at ISO6400 and ISO12800  Samsung NX1 crop sensor is significantly less noisier than full frame Sony A7 sensor.
Good for Canon 7Dm2 owners - at ISO3200 and ISO6400 it also has less noise than FF 24mp Sony A7. And in general it is not far behind 5DM3 in noise performance. So Canon really did some improvements in 7Dm2 sensor technology.

http://www.dpreview.com/reviews/image-comparison?attr18=daylight&attr13_0=samsung_nx1&attr13_1=canon_eos7dii&attr13_2=canon_eos5dmkiii&attr13_3=sony_a7&attr15_0=raw&attr15_1=raw&attr15_2=raw&attr15_3=raw&attr16_0=3200&attr16_1=3200&attr16_2=3200&attr16_3=3200&normalization=full&widget=1&x=-0.3743389872393283&y=0.5096437954588491

13
I think this technology could make it possible to create affordable compact mirrorless medium format camera using optimized pancake lenses which would require less optical elements in design (e.g. 3-4) elements instead of typical 10 -17 elements). So lens could be much smaller and cheaper which would result in significant cost reduction of overall medium format system. This could be breakthrough in medium format cameras design and could allow MF to take more market share from FF DSLRs

14
Lenses / Re: Canon TS-E 17mm f/4L Tilt-Shift Lens - your thoughts?
« on: April 10, 2014, 05:03:37 PM »
Though TS-E is perfect optical system it need to be used with care.
I've seen many images with TS-E when it was not used properly.
It like working  with HDR - if not use carefully than it easy to overcook.
Some people are overcooking prospective correction with lens shift  and make vertical lines  on image strictly vertical  and parallel to the picture sides and this breaks overall image prospective  - image does not look natural   - as if turned upside down.
When you look at the image upper side looks bigger than down side though they are equal .
This is optical illusion  because our mind uses it's own prospective correction .
So it is important  always to leave some prospective proportions on the  image and do not do it fully square

15
Lenses / Re: Canon TS-E 17mm f/4L Tilt-Shift Lens - your thoughts?
« on: April 10, 2014, 04:48:48 PM »
Neutral, that second crop is downright scary.  Wow, insane detail.  Very, very impressed with your work, technical skills, and the Sony+TS-E 17 combo results.

+1 on privatebydesign's comments, TS-E require a completely different style of shooting.  LiveView has made it a lot easier, but it's no substitute for the 16-35 unless you're using the 16-35 like a TS (i.e. architecture and other slow, deliberate type work).

Such kind of combo  TS-E 17+ Sony a7R  combo was always my dream.
Now it became reality. Thanks both to Canon and Sony.

 As for 16-35 I do not think it is really needed for anyone  who  already has Canon EF 24-70 f/2.8 M2 which is also perfect on both bodies.   
Canon EF 24-70 f/2.8 M2  is enough for most of short zoom needs, and  TS-E 17 is perfect complement to that when high quality wide angle shot is required. And both work perfectly on 1DX and a7R.
Here are also some image samples of   Canon EF 24-70 f/2.8 M2 on a7R and 1Dx: 
 http://www.canonrumors.com/forum/index.php?topic=20321.0

  With this set  EF 24-70 f/2.8 M2  and TS-E 17 I  really no need  16-35 any more  and to my view  IQ from 16-35 is far below  compared with IQ from EF 24-70 f/2.8 M2 and TS-E 17  even on 1Dx body let alone a7R, so I was never  fully satisfied with 16-35

Pages: [1] 2 3 ... 8