perfectRAW 0.6 alpha

I have spent near half an hour translating this info Vergüenza, so I hope you will find it useful. I am afraid that my English is quite rusted.

I know the are some English readers following us in OjoDigital forums. I want to say hello to all of them, and thanks a lot for your interest.

We will try to post in both languages in this page (in a very early stage of development, as you can see).

Just remember, perfectRAW is free, and will be donationware in the future. Future versions of perfectRAW will remain free.


In the bottom attachment you have perfectRAW 0.6 which includes this new functionality (relative to perfectRAW 0.5 alpha 2). It has been 100% compiled with MSVC (but may be compiled with gcc based compilers also):

  • AFD interpolation.
  • Luminance extraction for every interpolation algorithm (for AFD it is extracted inside interpolation, and for the rest after interpolation).
  • Raw noise attenuation for high ISO images, based in texture detection using AFD interpolation. This attenuation is only for high ISO images and can be used with any interpolation algorithm, because it is done before interpolation. It depends on two parameters, the first one (from 0 to 100) is the noisy pixels detection, the second one (from 0 to 100 again) is the level of attenuation for those pixels (0 none, 100 full). The idea is to reduce noise without removing texture, so you can get fine grain and less color blotches. Is very easy to reduce the noise of resulting image in Photoshop or whatever noise reduction software you use. If you use it in noise free images it will do almost nothing.
  • Small changes in the program GUI (better TAB behavior).

In some days I will upload 0.65 alpha version with this functionality added:

  • AFD interpolation parameters selection (at least the number of luminance extraction passes, now fixed in two, the optimum value).
  • AFD interpolation noise reduction based in median filter, filtering luminance and chrominace before mix. It is fully implemented, but I want to filter more the chrominance noise. Anyway it is not different of what you can get using your favorite noise reduction software.
  • Full ICC (sRGB is still fixed).
  • Fixing a couple of bugs I have identified.
  • Non optmized, SSE and SSE2 versions. Version 0.6 will work with SSE2 capable CPUs only (I am sorry, I am lack of time). Final version of perfectRAW will detect CPU capabilities at start up and will configure itself.

Known bugs:

  • TAB behavior is still a bit irregular. When in full screen top of the image is cropped (a mistery for me). TAB can not be used for moving through GUI controls.
  • Raw selection form appears on top after raw is selected.
  • Jacques Desmis (who is making a great work in quality testing perfectRAW/dcraw) has reported a problem with gamma calculation. He has found a bug in the gamma LUT calculation. I will fix it in the next upload.
  • You have to manually install the ttf font inside Windows' font directory. Otherwise frontend looks strange. This is not really a bug, so don't report it, please!
  • There is no error control at all. If you try to load an invalid raw program will crash!
  • Coffin's raw reduction (wavelets based) is not working in the high lights. It is already fixed in my code and will work in next upload.
  • Raw files from Fuji cameras still are rotated by 45 degree (I have fixed this a couple of times! Vergüenza).
  • Noise attenuation is producing artifacts in the borders of the image (sorry it is easy to fix, but I did not notice it until I upload this version).
  • Noise attenuation is producing artifacts in the image borders (easy to fix... I just didn't notice!).

So, what can you expect for 1.0 version?

  • I am leaving the GUI in ariznaf and Egon hands, so they can change it for the OpenGL and detacheables controls.
  • I will repair know bugs before any other work, including custom white balance and chromatic aberration.
  • I will rebuild actual code future plugin support (a must for perfectBLEND and perfectRAW 2.0 plugin system)

Finally, what can you expect from AFD interpolation against other interpolation algorithms (I will say the cons also):

  • If the speed/quality balance, PPG is the winner. I will use it by default, it is really fast!
  • In noise free images, AHD is more sharp and works great with edges. AFD detect textures better, specially in wovens, at is faster than AHD. If there is some noise present, AFD does not create artifacts, while AHD does. It is difficult to choise between them it there is no noise. Personaly, if there is some noise I do not like AHD at all, it makes lots of labyrithns. Without any noise I like AHD more, in fact, in some structures AHD performs better than AFD. Without noise, in nature images I will choose AFD, and AHD if there are many straight lines, letters, etc.
  • For high ISO images I will use AFD without doubt. Anyway you will notice the great improvement in AHD and high ISO images when using noise attenuation (based in AFD for noisy pixels detection).
  • For BW images I will use AFD again, unless the image is noise free. In that case you can choose AHD instead.
  • Para BN sin duda alguna AFD salvo que la imagen no tenga nada de ruido. En cuyo caso podéis optar por AHD.

Anyway, this is my personal opinion, you will have your own experience. In the meanwhile, I am working in my own interpolation algorithm, some kind of very enhanced AHD. I do not know if it will be released in the future, but my preliminary results are really promising...

A little comparison in image quality in high ISO images. This image has been obtanained with a Nikon D700 at ISO 25600 (downloaded from raws.fotosite.pl). In the last example I have postprocesed chrominance noise in Photoshop in a very fast way (so you can see some color smearing). I will implement chrominance noise removement in perfectRAW in the near future, and it will be better than that:

ACR/AHD/AFD comparison

AttachmentSize
perfectRAW 0.6 alpha.rar368.15 KB