Thursday, July 7, 2011

Arithmetical Endeavor

I can't believe it's Thursday already. I don't feel even being in this office and typing this! I'm fasting, true, but who said I can't sleep more? I'm just hoping the weather would stay stable as it is now...

Yet, who said you should adhere to the bad weather? Well, maybe from the safety side, yes it's better not to do anything with your camera without proper cautions taken care of. Well, I don't have such cautions, but I decided to work in the dust when we had a strike of a dust wave (as usual, every weekend). I didn't do much with my VR-head for some time now so I've decided to go and try it out there, in the yard of my own home...

Still a Home
The tone-mapping of this HDR panorama got me by surprise. My aim was to find a yellowish hue for the image but when tone-mapping I found out that however I change the slides it just turns out with some shade of magenta. Even Color Balance in Photoshop did not correct much of it. Not planning to do it as a QTVR, so I've decided to cut out the bottom and some of the top instead of cloning out as usual. I'm studying now more methods for easier Nadir shooting. I have to admit though that I was lazy here. I needed to do more of "selective" tone-mapping; like tone-mapping two versions of the HDR and then overlapping them and do some masking between the two. The level of black is exaggerated to add some melodrama and melancholy to the scene by the fact that it's a dust wave accumulating on the ground (taken at noon, the dust wave started around morning).
Not one of my favorites I have to say, anyway I had to do one more task with this panorama; tunneling. This time though, I had to go further before tone-mapping and do some edition in Photoshop to ensure the yellowish hue for the whole panorama and avoid that magenta look and shade. Some cropping as well was used just to remove a remnant of the tripod handle in the stitched image.

Being a Grain of Sand

The title was inspired by Simply Red's song Say You Love Me, because when I saw such a view after the final adjustments, I couldn't help but to feel tiny compared to the structure (which is supposedly to be my home!). Speaking of yellow of course doesn't mean to make it bright (I hate yellow), but a dark golden hue is fine. I think it's attractive to the eye.
This simple experiment now is pushing me further to think more about abilities and capabilities in panoramas venture. I'm quite convinced now that my tripods are not useful to this VR-head at all. Specially that the handle of the tripod would show up in the panoramas (and it is large and needs to be cloned out carefully). One of the methods for shooting the nadir might be useful in compensating this and dropping out the appearance of this handle in the panorama (by using Point of View optimizing in some detailed work). I hve to try this out one day.
Beside the tripods, I've realized that I can subtract a vertical panorama out of a full panorama! However, it's not going to be an easy task and if I'm aiming for a vertical panorama originally, then I better do it that way. The thing is, I never tried doing a vertical panorama with my VR-head, and to do so I need to flip the Vr-head (L-shaped) to make it horizontal, so the camera would rotate in a circle vertically. Not an easy task with such head mounted on such tripods I have already. This is one of the reasons to think about a new specialized tripod with exchangeable head, if possible.
However, beside this technical observation, I'm not somehow aware of the necessity to stabilize my work with panoramas in a coherent way; that is to make my procedures with every panorama I tend to do more regular and similar. Such regularization would make my work easier on computer in case I needed to extract a file from the whole set for any purpose, like that for a vertical panorama. Making a mental note now, I've decided to:

  • In each row, I will be taking 13 angles (0 degrees to 0/360 degrees) instead of 12. Used to do a mix between the two method, but I have to fix the number of angles in each row to this. One extra shot might be useful later on.
  • Despite the difficulties that I will encounter, I need to take a row at 45 degrees up and down. In the above panorama of my home, I tried to take two rows down at 25 and 50 degrees, hoping that would help to clone out the legs of the tripod naturally, but I was wrong. We are back to the old ways.
  • Two zenith and one nadir shots will be taken, each perpendicular to each other. This would help me out later on in case I need to extract a vertical panorama from the full panorama. Usual angles are 0 and 90. Complications expected for nadir.
  • Shoot the upper 45 degrees first, then 45 degrees below second. This is to keep the numbering of the slides unique for every panorama I would have to go through, e.g. slide #14 will always point to the shot taken at 0 degrees tilting 45 degrees upward (0,45).
  • Always shoot Zenith and Nadir at the end. Zenith first, and Nadir second. Nadir though is still to be considered as I'm trying to find ways to overcome this hard angle.

This said, it leaves me with some calculations to do to know where I stand and what I'm needing here, excluding the fact that I always shoot for HDR panoramas, I will count the number of angles as single images:

  • 13 angles X 3 rows (0,45,-45) = 39 shots.
  • 2 Zenith + 1 Nadir (might be 2 with complications) = 3 (or 4).
  • Total: 42 shots (or 43).
  • Slide numbers for the 4 basic directions would be (number,angle of rotation): [(1,0),(4,90),(7,180),(10,270),(13,360)] @ 0 degrees vertical. [(14,0),(17,90),(20,180),(23,270),(26,360)] @ 45 degrees upward. [(27,0),(30,90),(33,180),(36,270),(39,360)] @ 45 degrees downward.
  • Zenith will be on slides 40 and 41. Nadir will be on slide 42 and beyond.

I hope this systematic division of the panorama procedures would make it easier later on to solve problems as well whenever they occur before stitching (e.g. adjusting the exposure with respect to the surrounding slides).

  • Ardeaglais Cormaic Into Play:
Speaking of the vertical panoramas above, I've had a try with one of the first panoramas I've taken back in Ireland, in Cashel: Ardeaglais Cormaic, or Cormac's Cathedral, situated on the top of Cashel Rock hill, along with the castle of Cashel Rock.
The reason I chose this panorama specifically to work with and try to extract a vertical panorama from, is the dome that decorated the zenith in that place. I have to admit though I didn't situate myself exactly in the center under the dome but it was good enough to show the details of this dome. It would look great, I thought, to find the pillars that hold the dome in a weird shape going up and down and connected to two grounds; up and below.
Several difficulties arose. First, I had to re-merge the images into HDR, because I've deleted the previous files that were done for this place (to save place). Picking the required slides for this vertical panorama from the whole set was relatively an easy task (but a matter of memorizing the place in mind and picking the required ones). At this point, the matter of having a vertical panorama done directly at the place, and with the camera moving up and down in a circle, showed evidently. If I was do it again, I would set the camera in landscape orientation and go from horizon to horizon in almost a full circle (excluding the nadir point). This way, with a fisheye lens, I would have a wide-enough view to stitch over.
Our case here was different. The full panorama, and because of the VR-head, was taken with the camera being in portrait position, and because of this I had to consider 3 shots for every elevation or level. The total slides needed for this vertical panorama totaled 15: 3 at 0 degrees, 3 at 45 degrees, 3 at zenith (zenith + 2 on its sides), 3 at 135 (or -45) degrees, and finally 3 at 180 (or 0 backward) degrees. Now imagine we do this landscape orientation, we would probably need 5 slides only considering the wide (horizontal) view of the fisheye lens.
Second hardship appeared shortly after, when I realized that I can't flip certain slides in PTGui and make them stitch upside down. Maybe there is away and I don't know of it yet, but I didn't waste time here trying to figure out. I realized the best way here is to flip the slides themselves (i.e. open them in Photoshop, turn them 180 degrees upside down, and save them into a new file). Mainly, the files that were causing the problem were 6 slides that come after the zenith level (i.e. from 135 degrees and down to 180).
Third problem occurred when, after stitching, there were some high level of artifacts in the final result. Hot pixels (weird colored pixels in the dark areas of the panorama) and other splashes of colors. I had to go back to PTGui and tried to change the settings I've used to do Exposure Optimization in the previous stitch, and after that I've saved the file in OpenEXR format (.EXR) instead of Radiance format (.HDR). The problem was greatly with some need of fixes still, but it was much better than previously done. The thing is, I don't know what solved the problem really; the file format, or the change of exposure optimization settings? I've noticed some differences before in quality and other parameters between the two formats for HDR files. I won't be surprised if changing the format from HDR to EXR did indeed solve the problem. Finally the result...

Suas agus Síos (up and down)

There were of course some adjustments as usual and some blue strikes that I had to put down by Hue/Saturation adjustment layers. Before tone-mapping as well, I went on fixing the white spaces between the arches to give out some details before the process of tone-mapping, by changing exposure and gamma values. Over with this project, now I would think of other panoramas and try to predict their outcome in vertical panoramas like this.

I'm trying now to understand the concept of exposures more, specially for night photography and long exposures. Along with that, I'm delving more into the concept of metering as it is an essential part of the whole process for long exposures and night photography. I even got myself a new calculator to keep it up in my camera's backpack just in case I'd need it to estimate relative stops and exposures. I don't know how it was hidden from me all this time but the concept was simple all the time: a "stop" is a factor of 2. Increasing stop meaning doubling the exposure, and putting it down then we are halving the exposure. Then, a simple formula can be used in a certain manner to have a long exposure (with aid of metering), and thanks to the many websites that were filled with tips (too many to mention here):

T = S * 2x

where T is the time needed to get the same exposure at metering, and S is the shutter speed (time, in seconds) at metering, and x is the number of stops in play. Now depends on the situation, the formula might change for something else other than the shutter speed, but here I'm sticking myself to metering the scene in Av mode (thus knowing the shutter speed needed to achieve 0EV level). We would have to add several factors in case we are changing the stops in more than one way, like changing the ISO and the f-number (aperture), as well as the shutter speed itself.
Now, to my experiment, and here I've been changing the stops simply by adding ND filters. The experiment was simply to shoot a light bulb behind the window bars, and adding water sprinkles to make some sort of halo around the light source. In this case, we have to do a reverse engineering of some sort. Not exactly, but simply use the formula the other way around: I want to know how many stops I need to put down to make the shutter open for 5 seconds; long enough to sprinkle water in front of the camera.

Behind The Bars

With simple algebra, we fix the time to 5 seconds, and we have the shutter speed known from metering the light bulb, and what is left was the number of stops only to know what ND filters to use.

log2(T/S) = x

or in case the calculator can't calculate LOG for bases other than 10, then it would be:

log(T/S) ÷ log(2) = x

Using this formula, the result was 3 stops. Thus, I need to use ND8 filter, which reduces the light by 3 stops. Bear in mind that there will be fractions, so approximations are in use here. The result was as I liked, all I had to do is click the shutter speed (with 10 seconds timer to give me more time sprinkling water in front of the camera). I tried to shoot normal HDR for the scene but merging into HDR was problematic with the noise level, despite the nice blue shades in the shadows that showed after changing the color temperature, when merging, to fluorescent. I gave out the idea of HDR for this scene and I'm satisfied with what I achieved with water sprinkling on the scene. I have to say that water seems to increase the the luminance of the scene a bit, and probably it will be a dramatic increase if it was not sprinkling the water!

Weekend is here, and I'm so frustrated at what to do next. So much in mind and so less in power! I have to burn this mind for some photography ideas. I'm planning to order a new set of books for now. Would be good to spend time at work reading some of those. Not sure yet about the books I wanted, but surely one them would be about Night Photography.

Uh Oh

No comments:

Post a Comment