Page 1 of 1

Mt Shuksan & Picture Lake

PostPosted: Wed Dec 21, 2011 4:41 pm
by den
Just for fun... and exploring virtual exposures from a single raw file, I:

(1) created a 'dcraw v9.11' desk top application link to SNS-HDR's [could be either Lite, Home, or Pro v1.4.5.2] 'dcraw.exe' located in SNS-HDR's Program Files folder...

(2) dragged and dropped a single Canon EOS-350D CR2 raw file on the 'dcraw' link with the options: -v -n 35 -C 0.99890 0.99945 -H 2 -w -o 1 -q 3 -m 3 -W -b 3.5 -g 2.4 12.92 -6 -T

The 'dcraw' options include: wavelet noise reduction; chromatic aberration reduction; highlight white point blending; camera color balance; sRGB colors; AHD interpolation; chroma noise reduction; fixed white and brightness levels; 2.4 gamma with 12.92 slope; and 16 bit tiff output. The chromatic aberration reduction and fixed white/brightness values were specific for this image.

(3) using the resulting (2) image as a base exposure image, -2, -1, +1, and +2 EV equivalent image versions were virtually created using the RGB BrightnessCurve transform of Picture Window Pro...

(4) the 5 image versions were dragged and dropped onto SNS-HDR's desk top link [could be Lite with a Default preset option, or Home/Pro with a Default preset] where both alignment and de-ghosting were dis-abled... for image version blending and tone mapping. No other SNS-HDR editing was performed...

...and...

(5) using the resulting(4) image and Picture Window Pro v6.0.8, the image was edited to add more deep shadow and high highlight detail; slightly increased saturation of the sky; corrected for lens distortion; and then cropped and re-sized as shown in the following illustration which also includes the base/virtual exposure image set:
dcraw9-11_SNS-HDRv1452_PWP608_Mt_Shuksan_110824.jpg
dcraw9-11_SNS-HDRv1452_PWP608_Mt_Shuksan_110824.jpg (311.18 KiB) Viewed 14948 times

The full-sized final image version [3456x2304 pixels] is remarkably free of objectionable luma/chroma/interpolation noise artifacts when viewed at a 1:1 resolution without having done any noise reduction, yet the sharpness seems to be suitable 'as is' for monitor/internet browser anti-aliased display.

The automatic image blending and tone mapping, free of objectionable halos/aura of SNS-HDR is truly remarkable.

Thank you, Sebastian for making SNS-HDR available!!!...

Re: Mt Shuksan & Picture Lake

PostPosted: Thu Dec 22, 2011 3:44 am
by mgg
Den,

As far as I can tell the base image is not blown at all, so what is the benefit of the two darker versions?

Last month I made a series of spherical panoramas in our home and I needed the full -3 to +3 EV range to cope with daylight through the windows and dark corners in the rooms. At the end of the process I tried SNS-HDR using only one exposure from the group of seven, the lightest exposure that was not blown at all. I could see no significant difference between it and that produced by using SNS-HDR on the full set of seven. Have you tried a similar test?
I no longer have the bracketed groups so I can't repeat that particular experiment, Depending on your response it might be worth shooting another set.

Mike.

Re: Mt Shuksan & Picture Lake

PostPosted: Thu Dec 22, 2011 5:40 pm
by John 2
mgg, I agree with you that there is no point in applying SNS to an image that is basically OK and even the base image that I used for the Church interior (see below and I must remembr to switch off my spot metering)) could have been rescued by careful selective adjustment in Photoshop. The benefit to me of SNS in that situation is that selective adjustment would have taken a considerable time to achieve and I would still have had the problem of "hiding the joins". Whereas, creating a number of Raw images with progressively varying exposures to cover the range, was a few minutes work and SNS took care of the "joins" for me. The result is there to be seen.

Re: Mt Shuksan & Picture Lake

PostPosted: Thu Dec 22, 2011 7:31 pm
by mgg
John,
I'm afraid that I expressed myself badly. I do appreciate the value of SNS-HDR melding a number of differently exposed images together. Further, I see advantage in using it on single exposures.
The point I was trying to make is this. If you take a bracketed set such as this:
1 very dark, nothing blown
2. not quite so dark, nothing blown
3. a bit dark, nothing blown
4. brightish, some blown areas
5. bright, a lot of blown areas but not totally blown.

then I fully appreciate the benefit of using images 3,4 & 5 in order to get detail from the shadow, but I don't see the benefit of using images 1 and 2.
There is probably not a down side to using them - but is there a downside to NOT using them?
I admit that this is a somewhat theoretical point, but I'm asking in case there is a value in using the images 1 & 2 above that I'm missing.
Mike.

Re: Mt Shuksan & Picture Lake

PostPosted: Fri Dec 23, 2011 5:25 am
by den
There is a demonstratable and possible advantage to 'SNS-HDR image blend/tone map' either (1) virtually created balanced below/above exposures of a base exposure for a captured scene whose dynamic range is within a camera sensor's range or (2) multiple balanced below/above exposures of a base exposure using camera settings for those scenes beyond the sensor's range... ...by analyzing histograms.

For the image example in this thread, the scene's dynamic range was within the camera sensor's range so only a single capture was taken biased not to clip sky/cloud/snow highlights... ...and the below/above exposures virtually created.

Illustration 1: the base image, its Luminoisty, and Luminoisty histogram. Note that the majority of the histogram distribution occurs below mid-tone, the 50% [128] tone and has two main peaks with a histogram minimum approximately at the 25% [64] tone.
1-BaseImg_Luminoisity_Histogram.jpg
1-BaseImg_Luminoisity_Histogram.jpg (44.38 KiB) Viewed 14909 times

Illustration 2: the base image processed with SNS-HDR Lite-Default, the resulting Luminoisty, and Luminoisty histogram. Note that now there are four peaks more evenly distributed from 0 to 100% [0 to 255] tones with minimums around 15% [38] tone, 35% [90] tone, and 75% [191] tone.
2-BaseImgHDRLite-Default_Luminoisty_Histogram.jpg
2-BaseImgHDRLite-Default_Luminoisty_Histogram.jpg (57.05 KiB) Viewed 14909 times

Illustration 3: the base image including the -/+ EV image versions processed with SNS-HDR Lite-Default, the resulting Luminoisty, and Luminoisty histogram. Note that the four peaks are even more distributed and equalized from 0 to 100% [0 to 255] tones with minimums around 25% [64] tone, 50% [128] tone, and 75% [191] tone.
3-BaseImg+EVImgsHDRLite-Default_Luminoisty_Histogram.jpg
3-BaseImg+EVImgsHDRLite-Default_Luminoisty_Histogram.jpg (59.57 KiB) Viewed 14909 times

While Illustration 3 may be technically correct, does it appeal to the 'artist' within?

Perhaps not for everyone or every image... ...just exploring SNS-HDR possibilities...

Re: Mt Shuksan & Picture Lake

PostPosted: Fri Dec 23, 2011 6:46 pm
by mgg
Den,
What a neat way of illustrating the point.
I'm left wondering what the histogram would look like using only the base image and the two brighter shots...
Mike.

Re: Mt Shuksan & Picture Lake

PostPosted: Sat Dec 24, 2011 3:03 pm
by den
Mike... will respond after holiday weekend commitments!!!

Seasons greetings to everyone!!!

...den...

Re: Mt Shuksan & Picture Lake

PostPosted: Sat Dec 24, 2011 8:02 pm
by mgg
Den - and all the best to you and yours.

Mike.

Re: Mt Shuksan & Picture Lake

PostPosted: Fri Dec 30, 2011 3:02 pm
by den
Mike... here is an Illustration 4...

Illustration 4: the base image including the +1 and +2 EV image versions processed with SNS-HDR Lite-Default, the resulting Luminoisty, and Luminoisty histogram. Note that the histogram minimums tend to be progressively shifted to the right resulting in an image version that has increased mid-tone and highlight brightness and less overall contrast.
4-BaseImg+1EV+2EVImgsHDRLite-Default_Luminoisty_Histogram.jpg
4-BaseImg+1EV+2EVImgsHDRLite-Default_Luminoisty_Histogram.jpg (59.63 KiB) Viewed 14841 times

Please visit: http://www.ncplus.net/~birchbay/SNS-HDR/example01/index.htm for higher resolution illustrations and Mouse On/Off comparisons.

Re: Mt Shuksan & Picture Lake

PostPosted: Sat Dec 31, 2011 6:31 am
by mgg
Interesting and convincing. Perhaps demonstrating that the algorithms are biased towards the assumption that the -1 and -2 EV images will be present - as they would likely be in a 'normal' set of bracketed exposures.
Fascinating.
Mike.