01. I understand that Hugin [and various other 'Pano' pieces of software] has not been built with my particular image composition requirements in mind. I don't even own a digital camera, so you will have to forgive me if I do not use the correct terminology as I try to describe my a.) requirements, b.) explain where I am stuck at the moment, and c.) ask for suggestions on how to get the job done.
02. My requirement is to 'stitch' together a collection of over 100 separate tif files. The files average about 52MBs each. I mention that up front only to indicate that one of the problems I have had in trying to solve this problem with other software [Photoshop CS3, CS4, CS5; PtGui, and ICE] is the size of the inputs and ultimate output. So, while I expect that Hugin will have less trouble that what I have tried so far, I am also a realist, and want to say that if it is not realistic to solve the problem [without asking me to upgrade to a 64- bit OS and 8GBs of RAM] on my pretty decent XP workstation [Quad core, 4GBs RAM, vintage 2009], I am quite ready to accept cutting the number of patches up to about half that number. In other words, while pure size is part and parcel of my problem, I don't want to get hung up on that. I need you help in correctly configuring/controlling the workflow even if we just talk about a simple test case of a 4x4 grid. 03. So, on to the nature of the images and the reason for the title of this post. The images are a hand-drawn perspective of St. Louis in the decade after the civil war. The Library of Congress has made this phenomenal artwork available on the Internet using MrSID technology that works pretty well for the average home computer user who comes on line to take a peek. Since the 'mosaic' has many of the characteristics of a large scanner project, you might think that the correct way to approach this in Hugin would be to follow directions that are found [here and elsewhere] covering the right way to work if the folks at the LoC had worked from some huge original, and then tried to put it on a bed scanner to break it up into 100 pieces. That is not the case. Rather the case is that the original artists produced 100 separate drawings, conveniently including a small perimeter overlap between adjacent parts of the map, but the perspective that pervades all of the patches is what you would have gotten had you been able to place a wonderful camera up on a tower about a mile away from the geography that was included in the city at that time [1874]. 04. So, before I try to describe what I have been able to do, and not do, with Hugin, I think it is appropriate to describe what I have been able to do with Photoshop. Using the what they call the Interactive Mode of working with this kind of mosaic, I have been able to manually arrange a 4x4 set of images. Now I do not profess to understand either their trigonometry or yours, all I can say is that the 'sort of onion skin' view that the interactive U/I provides, in most cases I can very precisely align the required images, and then allow its "Snap to Image" algorithm make a finer alignment. [In some cases, however, it works better to disable "Snap to Image" and rely on my old eye under a magnifying glass.] Then, having made all the positioning decisions in this manual way, continuing with their workflow, it eventually performs a 'blend' that automagically seems to get the right result from sets of pixels which are not the same. They are not precisely the same in terms of alignment, they are, in some cases, pretty dramatically different in terms of what I guess you would consider the 'exposure' of the two layers, and they are not even the same with regards to content. For example, one layer of a street may show only the curbs, but another will have the name of the street written down the center, and somehow the Photoshop blend gets it right. Now, I am obviously not into flame wars. I am not promoting Photoshop. Indeed, the very reason that I am here is that I have not been able to complete the task with Photoshop. With my hardware configuration, and lots and lots of tweeking, I cannot even get to the level of a 5x5 grid without having the batch process die [with not useful log or other diagnostic output to help troubleshoot]. The reason for carefully describing what seems to work is that I am sure that it provides the answer to how I am supposed to use corollary features in Hugin to solve the trigonometry. 05. So, here is where I am at this moment with Hugin. On the 3-step Assist panel, I have 6 images loaded. They are a proper 3x2 grid that Hugin cannot align -- none of the products has ever been able to correctly align based on whatever the normal heuristics are. The images are far too complex and have almost of of the photometric properties of a trip to Yosemite or the White House. I have arranged them using the Control Points panel. Owing to the superior characteristics of your U/I for such placement, my 28 control points are undoubtedly closer to exact that what the equivalent must have looked like to Photoshop when I visually aligned the patches. On the other hand, I guess there is a less subjective measure of 'goodness of fit' because the Assistant screen at this moment also editorializes that the 28 control points [four along each common edge, no attempt to provide diagonal control points for the very, very small corner overlaps] have "Mean error after optimization: 3.1 pixel, max: 7.7, Bad fit". and the Fast Panorama Preview's Layout panel portrays that by showing the layout schematically with three light-olive-green connecvtors and four orangish-reddish ones! That notwithstanding, on the Fast Panorama Preview's Projection panel the 66x34 Rectilinear display appears to be very good -- not perfect, not as perfect as my manual Photoshop equivalent, but still very good. Now I wish I could simply coerce the software into rendering a full-sized version of what is being previewed, but I cannot. I cannot, sadly, even find a way to zoom on the preview window in order to see what we have at a detailed level. So, let's make that specific question #1: Is there some means of using the preview capabilities to be able to look at parts of the composite at full resolution? 06. The best way I can explain how I have managed to get what almost appears to be a correct solution is to tell you what appears on the rest of U/I. Step 1 of Load Images says I finally chose a Normal (rectilinear) lens, and that I arrived at a focal length of 86.92mm [although I am pretty certain that when I was challenged for that information during the load step, I forced all the images to have the focal length of 1200mm, that being my wild guess as to what sort of lens in 1875 could have taken a picture of the relevant portion of St. Louis if there had been about a 300-foot tower somewhere a about a mile east of the Mississippi in Illinois.] On the other hand, if I switch to the Camera and Lens panel I see that my six samples are still selected and that the only metric is that they all have a hfov of 24, and now that I think of it, at some point in the wild try and fail do and undo tests, I did change my mind and decided to see what would result if I assumed that there and been a decent wide angle 244mm lens available in 1875! In any case, that gets me to the almost-impossible to comprehend Optimizer panel. It has my handiwork: First, all the Image Orientation check boxes were cleared, then all the X_Y_Z check boxes for the six images were set, then I went back and cleared the first patch [even though it is in the lower left corner of the 3x2, I treated it as 'home' -- at least that is what I think those check marks meant]. The nub of all of my rambling is that I had pretty well deciphered from whatever I could read at the PtBuit site, your site, and any of the others that are oriented to some wonderful DSLR, automated tripod gear, is that I absolutely have to tell your software that it is not to touch yar, pitch or roll. And with all due respect to whomever created the drop down for Optimize and wrote the notes to explain those options, there is nothing there that seems to tell me how or what to do. I am sure there is a community of folks from the 21st century who eat that right up, but not only was the artwork created about 150 years ago, I was created about 75 years ago, and I am still happily shooting Ektacrhome with manual lenses, so I admit to being the alien from the wrong time and space on all your wonderful algorithms. 07. Finally we come to the heart of the matter. In the state that my project is at this moment, the Step 2. Align button is enabled, but the Step 3 Create Panorama is not. Every variation of every test to click on align results in a less-worthwhile preview image. I guess I understand that you are adding to my manual control points, control point parameters based on other settings, and I don't know how to say"'NO, I only want positional transformation and then blend that mix. Don't try to another thing to improve on what you've got." In particular, while I thought I had it turned off, running another align/ optimize cycle causes the 'exposure' of each patch to be destroyed by some attempt to normalize whatever the dynamic differences are between the six samples. I get a patchwork in which half of the patches have been so bleached out that nothing of value is left -- albeit, I do seem to have found how to avoid warping, skewing, etc. 08. If it will help clarify what I can only poorly explain with words, I will try to upload screenshots of previews before and after being forced to press the Step 2 Align button so you can see what I am talking about. Otherwise, if you can just provide some specific step- by-step gestures that might get me some output ieven if it is a lost cause for me to ever truly understand that all your control parameters mean in order to be able to drive them correctly by myself. Thank you for taking the time to read through this, and thanks for a great piece of software even if I do not get through the workflow to a final result. -- You received this message because you are subscribed to the Google Groups "Hugin and other free panoramic software" group. A list of frequently asked questions is available at: http://wiki.panotools.org/Hugin_FAQ To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/hugin-ptx
