Shotster wrote:
That's not as bad as using JPEG, but still not as good as using raw.
What you need for an HDR is a lot of exposure information. If you plug in 8 JPEGs, all exposed 2 stops apart, you'll have an overabundance of exposure information for an HDR file.
The only reason I can imagine truly needing RAW for HDR is if your shadow and highlight details have a different color cast. For instance, shooting in a cathedral where the dark interior is lit by tungsten lights and the stained glass lit by sunlight. You could alter the shadow and highlight color temperatures independently that way.
DrPablo wrote:
What you need for an HDR is a lot of exposure information.
That depends on what you mean by "exposure information". The problem is in thinking that EV is the only thing that matters. You also need as much pixel information (bit depth) as possible if you want the most and finest control over the end result. And if you shoot JPEG, the camera has already modified and discarded much of the pixel data.
DrPablo wrote:
If you plug in 8 JPEGs, all exposed 2 stops apart, you'll have an overabundance of exposure information for an HDR file.
First of all, JPEGs are only 8-bits per channel (24-bit). You've therefore tossed a ton of data out the window even before merging the images and thus before any creative adjustments. Secondly (assuming these are JPEGs directly from the camera, as I can't imagine why someone would shoot raw, convert to JPEG and then to HDR), you're relying on the camera to process the raw data to create the images which you then process (merge) to create another image which you then process (creative adjustments) to create your final image. You simply cannot get results that are as good as they can be by starting out with less information. I'm not saying the result won't be acceptable or appealing to you or others, but you won't have the creative latitude you would have by capturing raw. Plus, 2 EV is sometimes too much of a jump in exposures between frames to yield the best HDR result in my experience. And of course, whether the final image is acceptable depends on a number of factors, not the least of which are the nature of the image and its intended purpose. If the image is intended to convey a lot of sharp detail and subtle tonal variation and will be output as a high quality fine art print, then using JPEGs will simply give less than optimal results. If all you do is share images at Web resolution through a Flickr account, then you might get by with JPEG.
Being the penny pincher I am, I often make tea with several used tea bags, but it's never quite as good as a cup made from a fresh bag. ;-)
DrPablo wrote:
The only reason I can imagine truly needing RAW for HDR is if your shadow and highlight details have a different color cast.
White balance is but one of the things over which you relinquish full control if you shoot JPEG. And of course, as you indicate, white balance can be set post-capture. Rarely do I concern myself with the white balance setting on my camera. I instead leave it at "auto" and season to taste during raw conversion.
Steve, I'm not saying that I recommend shooting in JPEG for an HDR merge. But the main reason for that is that a good HDR merge requires a lot of manipulation, especially in choosing the white point and doing your local transformation.
You're right that a JPEG will have less information per pixel than a 16-bit RAW-derived file. But the thing is, if you take 8 JPEGs and merge them in HDR, you're creating a huge 32-bit pixel in place of the shared pixels in the source JPEGs. A lot of the excess information inherent in the RAW file would be redundant given a lot of overlap between the JPEGs.
I hate to be a cynic, but this comes from reading a bit of Dan Margulis who claims that repeated testing has failed to ever show a benefit of 16-bit imaging. It's not that I believe him, per se, but his point makes one ask oneself "what is the evidence behind my assertions?" In other words, when can you show unequivocally a head-to-head HDR conversion where 8 JPEGs were clearly inferior to 8 RAW images from the same scene.
That said, I'm playing devil's advocate -- I still shoot RAW and edit in 16-bit and in LAB and ProPhoto RGB. But I wouldn't bet the house that it's making much of a difference.
I think that the curve illustrated in this window from the tutorial is a bad example. One can't argue with the final result, where Mahesh has produced an excellent image after all the processing, but this curve is probably not a good example to teach with:
This shows partial tone reversal in the highlights, i.e. the highest tones will actually become darker than the next highest tones. The slope goes flat and then negative on the right of the curve. The curves are not multipliers, they are transformers (transfer functions). If you create a flat line across the middle it does not leave the tones unaltered, it flattens them all out to middle gray (modulo some small variation due to the local adaptation algorithm).
Partial tone reversal is a specialty effect sometimes called solarization or partial solarization or the Sabattier effect (sometimes spelled Sabatier). Solarization was a total tone reversal resulting from a massive overexposure of a film negative. The Sabattier effect is most commonly known through the prints of Man Ray and is achieved by briefly exposing the B&W silver gelatin print to white light during development, often accomplished by turning on the white darkroom light for a second.
In Curves windows like this, the curve should be "monotonically increasing", meaning that the curve should never dip down but should get higher for each step along the bottom to the right. The slope would never go flat or negative and would always be positive.
One can accomplish things like "darkening the highlights" by moving the curve below a 45 degree slope in that area, without making the slope ever go flat or negative.
Another thing about the curve as shown is that there are no highlight values above medium light luminance because the curve peaks in that range. The later curve applied a little boost, but still not much.
It is an artist's choice to introduce partial reversal of tones or to reduce the luminance of highlights so they never reach full value, but those choices should be considered very carefully and I would advise not using them except in exceptional cases, and not as the default way of proceeding.
