PRESENTING RESULTS FOR GROUP 7
Group Members : Jordan Little, Tyler Weston, Matt Kurjanowicz
- Roles of Individuals
- Little
- Taking photo series. (Granada, Park)
- Chief Editor of Swiki.
- Weston
- Worked on the user interface of the program.
- Took some photo series with Little.
- Used the program to create some final results.
- Created some test images (Red, Green, Cyan, etc.)
- Created Coke-Can Series
- Worked on the Swiki.
- Kurjanowicz
- Worked on the merging portion of the program.
- Created simple series of images (Building Scape, Chess Pieces, etc.)
- Used the program to create final results.
- Worked on the Swiki.
Final Results ( PREFER SIZES of about 800x600 )
 |
| The above image is one of our best results. Our program generated this image to show enhanced light conditions. We have several other final images in our last update as well. |
What IMAGES were used ( PREFER SIZES of about 400x300)
This is but one generated image, please see G7-Final for more images.
DESCRIPTION
What were the goals?
Our goal was to create a program that could combine any number of images according to the conditions that we set. By conditions we mean the weights of the merged images. The program needed to accept constant weights, weights in a linear scale over time, weights in a logorithmic scale over time, and weights in any other arbitrary scale. Our initial intentions were to merge images over time - thus creating an image of a time when no image was taken. Pretty early on we realized that it could be used in interesting ways to create motion effects.
How stock footage was captured
- The basic test cases (solid colors) were generated using Photoshop.
- Images of the Chess Piece and Coke Can were captured by placing the object on a table and the camera on a tripod. Then images were taken without moving the tripod but moving the subject of the image.
- The time-lapse images were taken by placing the camera in a single location for a long period of time. Images were then captured in small intervals to facilitate the testing of our program.
Other Material that was used (acquired!)
Some images were taken at night of the Arc de Triumph at night. Andrew Knight and Matt Kurjanowicz took these pictures together, but the images were not used in this project.
Tools used and HOW?
- We wrote the program in the Java Programming Language, and used common tools to create the program (Netbeans, Vim, JavaC, Jar, etc.).
- We generated some test images with Photoshop. A new image was created and filled with a solid color and then saved as a jpeg.
- Automator and other Macintosh OS X system utilities were used to organize, resize, and in other ways manipulate images.
- iPhoto and Finder were used to generate contact-sheet type images of the series of our photos.
Code or scripting (include as an attachment OR link)
Please see Group 7 Final Code
What worked?
Our program was pretty successful at everything we attempted. We even found other uses for it which can be seen in the coke can series and chess piece series.
What was unexpected?
The original intent of our progam was just to work on time lapse of landscapes over time. We realized after working on the program for a while that we could do a poor man's motion blur. This turned out pretty well (see the Coke Can series), and by doing this we found out that we could remove objects in the foreground of images. This is akin to taking a median filter to a series of pictures.
What did NOT work?
Some of the times there were alignment issues with the program. Also, interesting subjects were often filled with people, and large time-lapse series were not possible for this reason. We could have taken more photos and removed the people (or made them less obvious) using a median filter. This could be percieved as not really being a problem, because in a sense people are part of the scene, and showing ghosts of people is in a way showing the subject over time. Both it and those viewing it are changing.
How would you fix it?
We chose subjects that were less popular and populated. We used a tripod to account for alignment issues. Also, we only combined images that were aligned correctly, and we used smaller series in order to eliminate alignment variation. We also could add image alignment to our program or create a script that will use another tool to align the images, then merge them. We could also take more pictures which would hopefully allow one to remove the variants from the invarients of the image using a median filter.
Thoughts for the future of this effort?
Our final results are interesting and our program has many practical uses. It can be used to combine images to change the lighting, enhance lighting conditions, or showing motion. Our program could also be modified to generate movies, such as of a can (or object) moving across a screen and showing motion blur, or morphing a scene over time.
|
