Thursday, August 18, 2011

A Bulughman In A Lonely Night!

Officially 31. Not much to do about it really, except of having a casual day to myself and sleep as much as I wanted. Well, that won't exceed 5 hours in best conditions. This week was slowly passing really, but I didn't have much to do with my camera, so maybe this is the reason why I felt it as slowly going. Anyway, I was pushing myself forward with my Ayvarith recording and finally recorded and finished Chapter 3 of Alexander's story. I could hardly stop laughing at myself while recording, so please don't mind the quality and the awkward presentation!
Ayvarith: here.
English: here.
Memorizing now these moments and all that work I had to do with this story makes me smile a bit. I've totally forgot about the other conlangs that I've initiated but totally forgot about them, specially the one named Bulughman.

Beside this recording, which took on several days actually and not a single one day, I took the chance to use the full moon for my benefit and have a panorama made from the roof top. Almost a complete panorama and not only a 360 panorama. Not so satisfied about it but it was an experience that I liked indeed.

I. Outtake #1!
I can say it is my first night-time panorama for the time being. I had to use the full moon to my advantage before it disappears from the skies so I was willing to work whatever the temperature would be. Anyway, the temperature was not as bad as before (gone down to 30s C as told by forecasts, while it was in 40s last week!). It wasn't all perfect of course and the noise played a major factor in this, but well, it's a lesson. But the most significant thing about this panorama is the lengthened workflow, which probably was not a good thing to do after all...

A Lonely Night

The Workflow (The Drama):
  1. In the beginning and after settling with my tools on the roof top, I went on doing some metering just to estimate the time I would need. The necessity to work as fast as possible (the moon moves across the sky) made me trade off with raising the ISO to 1600. Since I'm going to take a bracketed exposure for HDR (-2EV,0EV,+2EV) it was essential to ensure that when I go on with auto-bracketing that the exposure time won't exceed 30 seconds. 30 seconds is the limit in Av mode. I've metered some spots across the sky and the building fixing the EV to +2 and raising the ISO to 1600 which was the minimum ISO I could have an exposure time lesser than 30 seconds (15 seconds it was). However, the camera also had other plans of itself! I tried to choose an ISO following the stops starting from ISO100, but not the ranges in between, that is: 100, 200, 400, 800, 1600, 3200, ...etc. Although there are some ISOs in between as 160, 250 and so on, but according to some articles that these would be more noisy because of the circuitry involved in increasing and amplifying the signal, and it is better to have an ISO within the range of full stops.
  2. To be on the safe side, I've used the hyperfocal principle again, and fixed the aperture to f/8. This aperture, however, did change by itself as the automated process was going on. There are some spots across the panorama that, apparently, required more than 30 seconds to achieve the +2EV exposure and I've noticed at some points that the f-number did change by itself within the camera to f/7.1 (i.e. larger aperture than f/8) to allow for more light in.
  3. The white balance is fixed according to effect and not in desire to achieve a true balance of colors. My previous work with the moon on the roof a week or two back gave me an impression that the blueness of the sky is better and more attractive than a normal regular one. The whole thing was set by using the LCD display (which is not a good judge after all but a good estimate let's say) and from there I've fixed the WB to the minimum value of 2500K. Blue.
  4. The work on the roof continued normally and my eyes adapted slowly to the darkness there but I needed my small LED light to check the scale on the rotating disk of the VR-head to make sure it reached a 0 degree after a full cycle. I didn't bother about a nadir but I did take 2 shots for zenith, which later proved unnecessary, specially that I have no intent for a QTVR.
  5. Now moving to the lab, i.e. my PC. I think my first mistake was made up here. In order to reduce the noise level which I was almost sure of to be unaccepted because of using ISO1600, I've directly converted all RAW files into TIFF 16-bit format (using Photoshop this time and not DPP). I made this move to run NeatImage on the TIFF files and clean the noise accordingly (and put little sharpness as well) but seems it was a wrong step. Reasons will come later.
  6. After running NeatImage and batch processing all the TIFF files for noise reduction, I ran Photomatix for batch HDR processing. I have now 41 OpenEXR files created from the "cleaned" TIFF files.
  7. In PTGui now and because of the nature of the HDR files and the night scene on the roof, and because so much sky was involved in some slides, there were many slides that PTGui could not identify any points for, hence the slides were astray and "orphaned", meaning not connected to any image at all. There I decided to go on with "model" method.
  8. The model method, as I call it, is simply to batch process the individual HDR files into JPGs or TIFFs and plugging them into PTGui. This way, usually, PTGui is able to find more control points more easily. We save the file, which contains all data about control points and other parameters and positions of the slides and apply this template to the HDR panorama we are working with. A lengthy procedure but at times, it is a must. I went on tone-mapping the HDR files in a batch process in Photomatix, but first I picked one of the darkest slides (and it was one of those that didn't connect to any other images) and tone-mapped it to make it brighter and clear in most of its regions. The settings used in this tone-mapping is saved to be used for the rest of the slides.
  9. To my ill luck, after tone-mapping all slides, PTGui still didn't identify some of these tone-mapped images and could not connect them to each other although many features in the scene are obvious. Seems however, and after a close check up, that the difference in the noise level between two adjacent slides makes it difficult for PTGui to put on control points between the two despite the "visual" apparent common features. Here, and at this point, I realized it was wrong to clean all the TIFFs from the early beginning, and later on, more confirmation to this thought appeared.
  10. I had to put my own control points here although it is a process that is tiresome and I really don't like it because, let's face it, a human being won't be as precise as a computer and needless to say someone like me who can barely concentrate for a long time. I'm not sure even how I'm typing this! Anyway, the addition was done and the file was saved, and the template or model was applied to the original HDR panorama in PTGui. Everything was organized almost with some few broken lines that I directly decided to fix in Photoshop instead of wasting time doing a game with blending priority.
  11. The HDR panorama is now stitched and time for the fixes and cuts in Photoshop. Beside the regular fixes, I had to make an adjustment to the exposure (putting some areas down and raising up some) but before doing all the "local" adjustment, a major (global) fix for the exposure was due by using the white dropper on the brightest possible area in the scene. This is to bring some balance and brightness to the scene (which later on resulted in a good HDR histogram relatively).
  12. Tone-mapping the adjusted HDR file proved more that cleaning the noise in the early beginning was a wrong thing to do. It became more obvious that there were batches of noises while some areas were relatively clean. Seems that I have to keep the noise reduction procedure at the every end of the workflow to make sure that I don't get such patches of noisy areas and clear ones. To over come this problem, and back in Photoshop after tone-mapping, I went doing my usual Median trick that I do usually on skin to eliminate rigid features. This though made me lose some interesting dots in the sky with the rest (stars do appear sometimes with a long exposure. Orion belt was even obvious in the images though it wasn't visible to the naked eye).

The histogram of the panorama in gaussian-like shape. Might not be always a good sign to have most of your exposure around 0EV, but it's good that there is no sudden cut in the graph itself.

II. Outtake #2!
Although the previous panorama was uploaded to stock sites (only one actually), I couldn't say I'm satisfied yet. For all the points and the drama I've mentioned above, this time I did make up my mind to make HDR files out of the RAW files directly, without any filtration and noise reduction. As expected, the matter of the noise became lot easier and lot "unique". This time and just for the sake of a change, I've made up a little planet projection, which proved quite interesting more than just a plain flat panorama!

The Lonely Planet

As long the panorama contains big portions of the sky with relatively interesting atmosphere in it, then I guess a little planet projection should always be considered. The moon here looks like a sun, making me wonder if all the "suns" we see in photographs are really, a sun!
The procedure here was the same as above with exception that the HDR files were composed directly of RAW files (in Photomatix v3.0). 
I've used the previously made model file to arrange the points and the stitching errors were the same in fact; the same broken lines, but at least the noise was relatively easier to handle here. When stitching the panorama, in PTGui you are given the option to choose an "interpolation" method and usually I don't change this and keep it by default to Lanczos. The interpolation is concerned with the pixels of the image as a way to predict locations and values of pixels on the image when your image is stretched or squeezed and so on, which is something normal in case we are talking about panoramas made out of several images to be stitched (and the images are distorted originally because they are taken with fisheye lens of course). I believe there is some relation between the weird color spots that I used to see in my previous panoramas and this point of the workflow. For the time being I think I will use the Bilinear or the Nearest Neighbor methods to stitch my future panoramas as these two are more regular in Photoshop itself, as Lanczos method is not an option in Photoshop.
Photoshop editing was essential as well to balance the exposure little bit (but the histogram this time was different and not balanced around 0EV like before), and after tone-mapping (which gave me a headache for memory problems and again, Photomatix v3.0 was better to solve this problem and not the new v4.0. Sometimes I wonder why did they make a new version!!) it was time then to plug it again into Photoshop.
NeatImage was run to clear the noise in general (specially the chromatic noise related to the high ISO of 1600). Yet, there are some portions of noise that were hard to remove by NeatImage and I had to sort it out with the Median trick, again. This type of noise apparently is related to the temperature of the atmosphere and they were more apparent and so obvious to note in the sky area, while in the occupied areas in the middle where the ground and the bricks rule, this type of noise needs a bit of concentration to notice. I've lost some stars in the sky, again, when I did the median trick but well, I smoothed the sky. The lower portions of the image specifically were hardly struck by such noise more than the upper portions.

I think the main reason for such stitching errors is the movement of the moon, which caused the shadows to move as well and with such movements in the scene it is natural to have stitching errors. Beside the moon, my own shadow as well would be a cause for such errors. In fact, despite the fact that I've cloned out most of my shadows with the camera and the tripod combination, I've discovered that are still some traces of my shadow somewhere. Go figure!

This is it for now, and I'm thinking now if there is a way to reduce this amount of noise caused by the heat. Seems next time I'm going to bring a hand-fan with me to cool down my camera while I work! Ah well... just kidding really. I need to look up new things to photograph hmm...

No comments:

Post a Comment