PRESENTING RESULTS FOR GROUP 4
Group Members : (Jon Schwartz & Joshua Hunt)
We worked cooperatively on all aspects of the project, so I will show no division of labor. We worked for the same amounts of time, so we did equal amounts of work. As to the actual tasks of the project, they can be divided into two simple parts: photographing and posing for the scenes and then editing the source images with autostitch, photoshop, and fireworks to produce the final image.
Final Results
|
What IMAGES were used
We had roughly 150 source images, which are on a CD that Professor Essa possesses. If you would really like me to post all 150 source images, I suppose I can, but it'd probably kill the swiki.
DESCRIPTION
What were the goals?
Our initial goal was to create a panoramic scene that had multiple instances of one person in it. Our project involved to include two people - ourselves - and include three different environments, which had to be merged together to form one image. Each of the three separate scenes was a panoramic stitched together using around four or five different angle each, and instances of the two people were added using the 150 source images.
How stock footage was captured
We used a Canon SD200 3.1 MP on a tripod set at maximum settings to capture each scene. One person would pose while the other took the picture. We would set the camera in one position, and then take all necessary pictures at that angle before changing the angle of the camera to capture the next part of a scene. This prevented alignment issues we had with earlier attempts.
Other Material that was used (acquired!)
None.
Tools used and HOW?
We used three programs for the final results:
- Adobe Photoshop and Macromedia Fireworks for the inserting people into either the background images for the panorama, or the already stitched panorama.
- AutoStitch to merge the background images into the panorama.
Code or scripting?
None.
What worked?
In general, everything worked to some extent. There were often problems, but that doesn't necessarily mean that it didn't work. We found that using Photoshop/Fireworks was the best means of inserting people into an image, and using a tripod made producing source images much easier. Both Autostitch and Photoshop's Photomerge function worked for stitching the panoramas together, but AutoStitch usually worked better. We also found that working when the sky is clear is better, despite having to deal with shadow issues. The tripod helped to resolve some of the shadow issues. Working when it's overcast caused problems with stitching the sky together.
What was unexpected?
Issues that we ran into that we did not expect had primarily to do with lighting conditions and rapidly changing weather. We would also occasionally have problems convincing AutoStitch that it wanted to piece together our source photos, although that wasn't entirely unexpected. We ran into alignment problems that we didn't anticipate early on. We also ran into problems with stitching an overcast sky together.
What did NOT work?
Initial passes at this project were done without a tripod, and that produced quite a few problems with alignment and getting AutoStitch to work, as well as creating difficulties with inserting people into the scene. We also attempted to do the project on a cloudy day to prevent issues with shadows, but found that the cloudy sky was more of an issue than the shadows. As I mentioned before, there were often problems that we didn't expect, mostly minor, but everything usually worked to some extent.
The only thing that did not work at all was our attempts with PhotoMontage, and that was due to alignment issues. Working with PhotoMontage after we had a tripod may have been a possibility, but by that time we had a system worked out for using Photoshop and Fireworks.
We also had difficulty with Autostitch merging photos that already had people inserted into them, which created a problem. The people we inserted would only be in one of the background images, so when Autostitch attempted to find common points in the images so it could merge them, the people we had inserted would cause Autostitch problems. Inserting people into a background that had already been merged with AutoStitch was difficult due to the distortion of the image, and when we inserted the people into the background images prior to stitching to them, AutoStitch couldn't put the images together.
How would you fix it?
Using a tripod fixed pretty much all of the alignment issues and made setting up the scenes much easier. We were able to insert people into the scenes with much more ease, and AutoStitch was able to stitch the images together better.
To solve the problem of AutoStitch not being able to work with already manipulated images, we had to plan out our scenes carefully. We would avoid putting people near the edges of the source images. So, majority of the people in our scenes were near the center of the source photos. When working with the images, we would insert everyone that was not near an edge in the source photos, and then merge the images together. We would then insert the people near the edges of the source photos into the stitched image. This was difficult due to the distortion of the image, but since we already had majority of the people in the image, it wasn't too difficult. The inserted people often had to be stretched and skewed to fit into the image appropriately, but we managed to accomplish this without too many problems in the final image.
Thoughts for the future of this effort?
Other details and DOCUMENTATION.
Progess Images:
|
|
