Much work was needed to produce a high-quality virtual tour of the sunroom. Here are some of the things that needed to be done and some of the problems that remain in the panorama:
Our equipment was calibrated to within a fraction of a millimeter to ensure that all images were taken from the same viewpoint.
Images of mirrors and glass in picture frames and windows can contain reflections that we may not want to show in the final panorama. In this panorama, we wanted to show a clear image of the Japanese print on the wall, but it contained strong reflections of other objects in the room, so we took a second image of the Japanese print under different lighting conditions and combined the two images when making the panorama. We also used a drawing program to replace the matte around the print in the frame with a solid color.
The images below show the print with reflections in the initial panorama, the print lighted to reduce reflections, and the print without reflections in the final panorama. You can click on them to view larger images.
The tripod is below the camera when we take the pictures for a panorama, so we need to remove it if we want to include views looking straight down in the final panorama. It is hard to align this image with the images taken using the tripod, and editing must be done to remove tripod legs from images tilting downward before they are used to make the final panorama. Because of these difficulties, and because the view directly below the camera often does not add much, it is common to either limit the downward viewing angle or place a circle with credit information at the bottom of panoramas.
Lenses are not perfect, and the wide-angle lenses used to make panoramas tend to have significant distortion. There are mathematical models of lens distortion, but the program that we use to "stitch" images into panoramas uses a very simple model, so we can improve the "stitching" by cropping the initial images to shift the image center that will be used in the "stitching" program's lens distortion model. We now use a custom program with a better distortion model to correct distortion in the initial images.
Wide-angle lenses let less light fall on the sensors in the corners of digital camera CCDs. As a result, the pixels in the corners of the resulting images are darkened and do not match the pixels in other images when they are "stitched" together.
We wrote a custom program to measure and correct this darkening in the initial images (that were cropped as described above).
A program is used to "stitch" images together to make an initial panorama, and then an image-edting application is used to produce the final panorama. 32 images were used to create the initial panorama.
If images do not overlap enough or do not contain enough detail, the "stitching" program may not be able determine how they should overlap, so they may need to be aligned by hand. Even if the images do overlap and contain enough detail to fit together, they may not align perfectly because the lens distortion is not completely corrected by the "stitching" application. Problems with image alignment become much more noticeable as the resolution of the images increases. For example, the alignment problem shown on the right would not be noticable in a low-resolution panorama.
After the images used to make the initial panorama are "stitched" together, the blending of the images can be changed to increase sharpness in areas where images do not align well.
This is easiest when a structure to be sharpened, like a window, is contained in a single image and surrounded by an area with little detail, like a solid wall.
There are limits to the amount of sharpening that can be done in large areas that are not contained in single images or in cases when lighting varies greatly between images.
Sunlight and other lighting sources vary as the images used to make the panorama are taken. This can cause problems that require correction when the panorama is assembled. In addition, breezes, even to soft breeze made when someone walks by a bush, can move plants outside so that they do not line up correctly when images are "stitched" together.
The image on the right shows how the lighting in the sunroom varied when different images in the panorama were taken.
When we took these images, there was some dust on the camera's digital sensor that left small dark spots in parts of the panorama that could be hard to correct. Since then, we've started taking test shots after changing lenses to find dust before it becomes a problem.
The programs used to display panoramas are fairly new, and they occasionally have problems working as expected. In the case of the sunroom panorama, the program that we use to display high-resolution panoramas added some blue lines in a few places. The company making the program does not know why the blue lines are added. We tried various settings to try to get rid of the blue lines, and we found settings that reduce them, but we almost gave up trying to get rid of the lines before we found a workaround that got rid of them. Now there are small dots at the very top and bottom that we can't get rid of, but at least the problem is limited to those two spots.
Apple Macintosh computers attempt to display images as they will look when printed, while other computers attempt to display images as they will look on a television. Because an image will look differently using these two display methods, we produce two sets of image files so that our images will look right on all computers. For more information about this difference, see Poynton's Gamma FAQ.
Click here to view the sunroom panorama.
Copyright © 2003 - 2004 Experience Images, all rights reserved.