Highlights in RAW Convertor

[Brownie127]

Can you help explain why my highlights are blowing when converted in RAW Convertor.

Basic settings: v1.9 & 1.8, XPPro, NEF Raw, sRGB, into PSD 16bit or TIFF 16 bit, all as shot, Postprocessing disabled, Normal conversion.

Highs with RGB values around 245 indicated in Main View become 255 when viewed in CS2 with colour sampler tool.

[DavidB]

Is there a noticeable shift in the histogram between the NEF file (or, strictly, speaking, the preview image in the NEF file, as the RAW data itself is not viewable) and the converted image? If the highlights are as high as 245 to start with, it won't take much of a shift during the conversion to blow them.

BB Pro uses DCRAW to convert NEF files, and, as with all RAW converters, results can vary between camera models. I have no direct experience of NEF conversion, as I'm a Canon user, but, from what Chris has told us, DCRAW results are better with the more recent models. In any event, 'as shot' will always be an approximation when the assumptions made by the camera maker and the conversion software writer are likely to be different. It's a pity that you don't have any equivalent of the 'combined' conversion available for Canon files, as I find that balances highlights and shadows rather well.

You could try using Auto Levels in post-processing to see if that helps. Alternatively, if you are able to use another RAW converter, try that. For instance, if you have a recent Photoshop product (CS3 or PSE 5 or 6), the more recent (4.x) versions of Camera RAW are a great improvement on their predecessors. I also rather like Capture One v4. Both of these have settings to preserve highlight detail; the one in C1 is very effective, but you tend to pay a price in increased noise, particularly in high-ISO images.

I suppose that the bottom line is that you need to experiment; I hope that these pointers help.

[Brownie127]

Thanks for provoking thoughts.

Yes, there is a marked difference in histograms, I accept BBpro views will be using an embeded mini jpeg within the NEF RAW.

This thought has made me turn to the 'Raw Convertor Previewer' and scan the new enlarged image (Thanks Chris) with the sampler. Yes, the highlights are already showing 255 levels using the sampler.

I presume the Raw Convertor Preview is the actual image and therefore I must conclude a need to set camera exposure compensation around -0.3EV to -0.7EV and test again. I am happy with this conclusion.

As for Raw Convertor AUTO levels, I believe this would only aggrevate the issue as clipping at anything other than 0.00% would push more whites to 255.

Alas, I do not intend to go to CS3 just for the convertor. I prefer to stick with CS2 and BBPro. I do have the Nikon own ViewNX software which includes raw convertor. If I can learn how to nullify NX Picture Controls changing the file to unknown levels it could be interesting.

[DavidB]

I presume the Raw Convertor Preview is the actual image and therefore I must conclude a need to set camera exposure compensation around -0.3EV to -0.7EV and test again. I am happy with this conclusion.There is a common misconception here. The embedded JPEG is a RAW conversion like any other (and, like any in-camera JPEG, has been produced under the time and processing constraints applicable to the camera). Arguably, a significant amount of data has already been lost in the production of this preview, and it will rarely represent the maximum information that can be extracted from the RAW image data, which, as I said in my last post, can never be viewed other than through some conversion process. Your NEF preview is 'original' only in the sense that it was produced using Nikon firmware, and you can reasonably expect that Nikon will understand the characteristics of their own camera sensor. But the process of producing a conversion in the camera will always involve some element of compromise.

In these circumstances, it is worth trying different RAW conversion settings before committing to a particular scheme of in camera exposure compensation, which itself may need to vary under different lighting conditions.


As for Raw Convertor AUTO levels, I believe this would only aggrevate the issue as clipping at anything other than 0.00% would push more whites to 255.This is exactly the situation that Auto levels is intended to avoid, in that its purpose is to shape and place the histogram to preserve the maximum amount of detail in shadows and highlights. There is a useful explanation in BB Pro Help.

My experience is that Auto levels works well where the picture is generally well exposed, and you need to deal with relatively small areas of clipping (shadow or highlight). Problem images tend to need the more targeted (or aggressive) tools found in the specialist converters and editors, and there is usually a price to be paid in, say, loss of contrast or increased noise for the recovery of shadow or highlight detail in these images. However, where the embedded preview is already OK, you can be confident that there are RAW conversion settings that will produce an even better image.


Alas, I do not intend to go to CS3 just for the converter. I prefer to stick with CS2 and BBPro. Very understandable. I decided some time ago that Photoshop proper is overkill for what I need, and that PSE (which I do not always upgrade) will fill the bill for me. I suspect that there are a fair number of Photoshop users out there who could come to the same conclusion, if only their pride would let them. In my opinion, BB Pro is far superior to Bridge as a workflow application, in any case.

[Chris Breeze]

Thanks for provoking thoughts.

Yes, there is a marked difference in histograms, I accept BBpro views will be using an embeded mini jpeg within the NEF RAW.
That's right. You can extract the preview JPEG using the "Extract embedded JPEGs" option. This is useful when you want JPEGs quickly for proofs etc. and saves having to shoot raw+JPEG.


This thought has made me turn to the 'Raw Convertor Previewer' and scan the new enlarged image (Thanks Chris) with the sampler. Yes, the highlights are already showing 255 levels using the sampler.

I presume the Raw Convertor Preview is the actual image and therefore I must conclude a need to set camera exposure compensation around -0.3EV to -0.7EV and test again. I am happy with this conclusion.
The preview is generated by converting the raw data but it uses a simplified, faster demosaicing algortihm and only performs an 8-bit conversion whereas the full raw conversion uses 16-bit data and a higher quality demosaicing algorithm. These simplifications don't really affect the quality of the preview but significantly reduce the update time.


As for Raw Convertor AUTO levels, I believe this would only aggrevate the issue as clipping at anything other than 0.00% would push more whites to 255.
Yes, auto levels won't improve highlights and if set too high will clip them. Levels are useful if the histogram tails off to zero at the ends and can be used to select the best part of the 14-bits of data and map it to 8-bits when writing 8-bit images. The raw conversion routines use 16-bit data for all the adjustments and only convert to 8-bits before writing the image file.


Alas, I do not intend to go to CS3 just for the convertor. I prefer to stick with CS2 and BBPro. I do have the Nikon own ViewNX software which includes raw convertor. If I can learn how to nullify NX Picture Controls changing the file to unknown levels it could be interesting.

[DavidB]

These simplifications don't really affect the quality of the preview but significantly reduce the update time ... Levels are useful if the histogram tails off to zero at the ends and can be used to select the best part of the 14-bits of data and map it to 8-bits when writing 8-bit images. I think that the point here is that the JPEG preview will be as good as if you had shot JPEG rather than RAW, but the RAW data may (and in my experience often will) give you a better image if the full potential of the available data can be used. Otherwise, there would be no point in shooting RAW.


Yes, auto levels won't improve highlights and if set too high will clip them.This surely has to be read against the background of what is said in the Help file:


When auto levels is selected the black and white points are represented as percentages ... Setting both the black and white point percentages to 0 ensures that no shadow or highlight detail is lost and the darkest pixel in the image is black and the lightest is white. In practice ... the percentages are normally set to a low value (e.g. 0.01%) rather than 0%.
In cases where Auto Levels can move pixels from 255 to a lower level (typically, as Chris says, those cases where the histogram tails off at the ends), it will reduce the degree to which highlights are blown. But if your histogram is bunched (or has a spike) at one or both ends, Auto Levels will not help. However, in such cases, the JPEG preview will often look pretty awful, and the original point was that the problem was evident in the RAW conversion but not in the preview.

[Brownie127]

Thanks for your comments Chris and David.

I decided to test and compare alternative RAW convertors. There seems to be some noticable differences on conversion results with respect clipping and chromatic fringing.

There is no doubt the Nikon ViewNX and Nikon CaptureNX2 results on NEF RAW are first class, but being versions 1.0 and 2.0 they lack the ability to introduce customisations when making conversions.