Monday, November 2, 2015

Philae : La facade du pylone dite cossiti [?]

Some cool image sites images:

Philae : La facade du pylone dite cossiti [?]
image sites
Image by New York Public Library
Digital ID: 81487. [186-188-]

Source: [Collection of photographs of Egypt and Nubia.] / A. Beato and others. (more info)

Repository: The New York Public Library. Photography Collection, Miriam and Ira D. Wallach Division of Art, Prints and Photographs.

See more information about this image and others at NYPL Digital Gallery.
Persistent URL:

Rights Info: No known copyright restrictions; may be subject to third party rights (for more information, click here)

A Night at the Museum
image sites
Image by Stuck in Customs
This is one of the final photos I was able to squeeze off before the security guards nabbed me. Arg! I wanted to get so many more… but you know, you know… it’s the same old story.

Anyway, if you ever visit Paris, you absolutely need to make it over here to the museum of natural history. It’s known to the locals, but not to visitors. All the other sites in Paris are great, of course… but if you have a penchant for science and a love of crazy interior design, this is the place for you!

- Trey Ratcliff

Read the entire post over at the Stuck in Customs blog.

A Slight Shift
image sites
Image by naggobot
Notice how the right side crane is a bit blurry. It's actually perfectly in focus. I do not know why it looks blurry, probably something to do with the girder structure. It's the perfect looking left side crane that moved during exposures.

Photo is from last winter and on the background is the construction site for Vantaa waste power plant. Now it is a bit unfair to talk about this place as Vantaa waste power plant as the collection area extends from this location over 150 km west all the way to Hanko. Yes, it's only west, not east.
Notice how the highlight of two lamps is almost perfect. I think I finally managed to get the grips of generating non magenta highlights to HDR:s. Note that this is probably not an issue if you are using latest LuminanceHDR.

You may experience magenta highlights if you use 16 bit tiff or raw images to make HDR's.It is possible to make HDR's from jpg but my experience is that the colours will be dull. I am unfamiliar with the improvements on RAW handling of Luminance HDR but on the very old 2.0.2 version I use it is tricky. My work flow is Rawstudio -> 16 bit Tiff -> HDR. This work flow produces good and vibrant colours but over saturated areas turn typically to magenta, sometimes green and fixing these manually in Gimp can be quite difficult.

Also on this image the lamps on the centre had been over saturated and after tone mapping the highlight was pure magenta. I rejected the HDR and started over a bit differently. Steps were

1. Generate 16 bit tiff images using Rawstudio
2. Cd to image folder
3. Open base exposure with Gmic. Just type gmic bracket_1.tif (bracket_1.tif = your image file name)
4. In the preview window move your cursor on top of the highlight. Make a note of the values shown by the preview window. These values are shown in a corner. It may read something like 65535,60244,65535. These values are for RGB channels and the green has not been over saturated. Find the highest peak value for green in the highlights and make a note of it.
5. Close preview window.
6. Type gmic bracket_1.tif -cut 0,highest_peak_value -normalize 0,65535 -type ushort -o braket_1_cut.tif (here highest_peak_value=a bit less than the highest peak you found from your image. If you had 60244 then maybe 60200 would be good. I have tried this only twice yet so I can not say for sure.)
7. Repeat step 6 for all brackets with the same highest_peak_value.
8. Generate HDR with Luminance
9. Tonemap

What happens in step 6? Lets break it down.
gmic: Calls gmic
bracket_1.tif: Tells the file name to read for gmic. At this point image is converted to 32 bit per pixel because gmic does all processing as 32 bit.
-cut 0,highest_peak_value: Cuts the image data so that values above highest_peak_value are thrown away. For highlights this means that magenta turns now to grey.
-normalize 0,65535: Stretches remaining data back to full 16 bit range. This turns our highlights back to pure white.
-type ushort: Changes the pixel data back to fit 16 bits
-o braket_1_cut.tif: Writes output to file braket_1_cut.tif

Few considerations
It is worth while noting that this method changes the colour balance of the whole image. In this case it was not much but in other cases it might be. If the method produces good highlights but bad white balance otherwise then make another image with out the cutting and use Gimp+masks to combine the two images. I also tried just normalizing the green channel and it worked too but affected the white balance more in this case. The command used was
gmic image.tif -split c -cut[1] 0,60000 -normalize[1] 0,65535 -append c -type ushort -o output.tif

For this image two different HDR's were produced, one with camera white balance and one with spot adjusted white balance. These were combined in gimp to produce the orange tinge of the lamps. In addition quite a lot of other masking and colour tint edits was made (for example the bluish border). Final image has a total of 12 layers, bottom two of these are the tonemapped LDR's made from two cut value HDR's.

For comparison of different raw developers see

I found out that it is possible to controllably produce excellent results with UFRaw by adjusting EV. Same method should work also on other raw developers but it is not possible to know the correct adjustment value from the information shown. UFRaw gives percentage values for the amount of overexposed pixels and it seems that this information can be used to set correct exposure compensation.

This photo is Creative Commons licensed so you can distribute it freely and you may use it for any noncommercial purpose as long as you attribute the author. CC license also means that you may not use, distribute, abuse or steal this image under any such license that you pay of, namely you may not use or distribute this image under Finnish Copyright Groups Digilupa or corresponding license.