We want to design a system that can help people become better photographers. The ideal application would be a "smart camera" that provides artistic suggestions directly in the viewfinder. The problem is, it is unrealistic to teach a computer to be an expert photographer. Fortunately, many photography techniques are simplified into easy-to-apply rules.
![]() |
![]() |
![]() |
| Balance | Rule-of-thirds | Spacing |
|
We consider three rules for measuring composition. Balance is defined by the proximity of the center of mass to the middle of the image. Rule-of-thirds represents distance to any thirds-line. Spacing relates to the distance between objects and the borders of the frame. |
Our hypothesis is that basic composition rules offer an improvement to the average person's photos. To test this hypothesis, we construct an image cropping experiment that approximates the choices made in photography.
We developed a program called CropTool (download) to facilitate image annotation. Written in C++ using the Qt toolkit, this application allows users to draw bounding boxes around salient elements of an image. The data is exported to a custom label format representing a vector of rectangles.
Collecting Computer DataWe developed a program called SmartCrop to automatically crop a labeled image. Written in C# using the .NET framework, this application uses a Monte Carlo algorithm to simulate potential crops. The candidates are scored and pruned based on a composition heuristic developed in concurrent research (see Panorama project).
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 |
|
|
|
| Human Crops (top) | Annotated Image (middle) | Smart Crops (bottom) |
UC Santa Cruz, Computer Graphics, Fall 2010