Update 1 for Group Seven
Group Members : Matt Kurjanowicz, Tyler Weston, Jordan Little
- Roles of Different team-mates:
- Kurjanowicz
- Begin work on the merging application.
- Gather rudimentary sample images.
- Weston
- Take lots of pictures over time of interesting subjects to use as input for the application.
- Help work on the application.
- Little
- Scout out interesting subjects to take time lapse photos of (Gehry's fish, Mies Pavillion, etc).
- Take pictures of those subjects.
- A Time line
- June 29: Progress so far
- Have begun working on the application to merge images taken over time.
- Have a small library of controlled sample images.
- July 5: Will show
- Have lots of pictures of different subjects completed.
- Begin looking at initial results.
- Select final subjects for interesting pictures.
- July 10: Will show
- Finish program.
- Finish compiling pictures.
- Analyze results.
- Problems to date (28 June 2005)
It is difficult with digital cameras to get an accurate set of sample images. The digital cameras try to adjust to keep the color of the image fairly consistent, but not from shot to shot. So a picture taken at 8:00 PM may not be what the eye sees, nor may a picture at 10:15 PM, but unfortunately the way the camera skews these images is (at least to us) unknowable. This may not matter because the generated image may be very close to how the camera would capture the image at that time anyways.
We have encountered a problem in realizing that our idea may not work. If you look at generating images of times that are unknown, some things must be taken into consideration: color, ambient light, and shadows. We believe that our idea (and thus application) will not deal with shadows correctly, but that it will deal with color and ambiant light levels correctly.
Also, some of the programming work has not gone as quickly as planned, as we hoped to have had a working demo by 29 June 2005 of a basic version of out idea.
- Results and or images used to date (use small images please!)
Here is the application that we're writing. It is far from being done, but it goes through all the motions correctly (but doesn't actually work when it comes to merging the images).
| Main Window (What the user sees when they start the application): |
 |
| Then the user adds images and enters the time that the images were taken: |
 |
| Then the user enters a time for the desired image and clicks "Merge!": |
 |
| ... And the current result: |
 |
We've also collected some very controlled sample input images:
| First Input Image (8:25:30 PM): | Second (9:44:44): | Third (10:01:51): | Fourth (10:10:20): |
 |  |  |  |
| (Almost) all of the input images in our current sequence (from 8:25:30 PM to 10:10:20 PM, at intervals no greater than 10 min): |
 |
| A sample of the recorded times for each image: |
 |
|
