Monday, March 17, 2014

Dual-ISO experiments

Some of the responses from my recent posts about Magic Lantern on the 5DMkIII, indicate that not all are convinced as to the value of dual-ISO; where, in a single image, you capture a wider dynamic range than is possible in a normal image. The ‘criticism’ is that post processing workflow is too complex and/or time consuming.
Well let’s look at that assertion.

My  5DMkIII is able to capture a DR of about 7Ev (rounding down a bit) and, for metering, ie clipping, I would say it was 6.5Ev. Thus, if the scene I was capturing was wider than this I would usually adopt a bracketing strategy; and throw these brackets at one of my HDR pieces of software, either tone mapping the brackets or using a fusion algorithm. Note all this this takes time, eg: first setting up the bracketing strategy, then taking the brackets, then ingesting into Lightroom, then exporting the brackets to, say, Photomatix, then bringing back the processed TIFF file into LR, then post processing the new (TIFF) ‘negative’

So what about an ML dual-ISO LR-based workflow, well it goes like this (some steps you only need to do once to set up the workflow): 

  • Ensure dual-ISO is running in ML and set, ie I use 1600 as the second ISO for about 3 Ev boost in the DR of the captured image, ie taking a single image capture to about 10 Ev, from about 7Ev;
  •  Switch to Live View (LV);
  • Survey the scene with the ML spot meter, ie highlight areas to dark areas (where you still wish to see detail); 
  • If the difference is more than about 10Ev, dual ISO will still produce areas of ‘blocked black’, but this may be artistically OK. If it is not, adopt an appropriate  bracketing strategy;
  • Use the ML spot meter to meter the scene by moving it to the highlight area you wish to protect and adjusting the exposure, usually shutter speed, until the spot meter reads just under 0Ev, ie it will be blown out at -0Ev (sic) thus you want to see -0.1Ev or so; 
  • Ensure your shutter speed is OK for handholding (or use a tripod or some other stabilizing strategy); 
  • Capture your dual-ISO single image;
  • Ingest into LR; 
  • Have cr2hdr.exe on your desktop (with dcraw.exe and Exiftool.exe – all downloadable from the ML site). I put all three .exes in a folder for easy access. Once you have these three pieces of software available, you don’t need to play around with them anymore; 
  • Export your dual-ISO .cr2 file, ie RAW, into another folder on your desktop or into the same one as the cr2hdr.exe; 
  • Switch to your desktop and drag the exported RAW on to the cr2hdr.exe. The .cr2 will be automatically processed and a .DNG file placed in the same folder as the .cr2; 
  • Go back to LR and import the DNG, which now becomes your ‘new’ negative – just like in the case of the tone-mapped or fused brackets;
  • Finally post process as normal.

The attached handheld dual-ISO was taken this evening to provide an illustration of the dual-ISO output. I estimated the base scene to have a DR of over 10Ev, ie window area to bookcase. Although not a winning image, it does show that a post-processed dual-ISO offers an alternative route to capturing wide DR scenes, ie an alternative strategy between non-dual-ISO single capture and bracketing.

Although the workflow above may look complicated, once you have been through it a few times it becomes very quick. Certainly not taking any more time than processing a set of (HDR) brackets.

Bottom line: for the Canon DSLR shooters the ML dual-ISO offers an interesting additional strategy to capturing high-DR scenes, especially if you are forced to use a single-image capture approach, eg because of subject movement.

No comments:

Post a Comment