Online Image Segmentation
Di Zhong and Shih-Fu Chang
Department of Electrical Engineering, Columbia University
Segment images and get the objects you want !
The applet demonstrates one basic region segmentation algorithm that we have developed for images and videos. The algorithm effectively combines color and edge information to obtain accurate region boundaries. Detail description about this algorithm can be found in our paper published in ICIP'98.
The applet is developed in JAVA using JDK 1.1.x. You need to have a web browser that supports JDK 1.1.x or above to run the system.
Please follow the following steps to use the system.
- Image URL : To provide the URL of an image that you want to segment. The image can be in one of the following formats: GIF, JPEG, PPM or PGM. Note that images larger than 480x400 will be downsampled before the segmentation. Click on "Image URL" button will show some demo images and the history.
- Segment Image : To segment the current image. The segmentation process may take A FEW MINUTES for large images. There are 3 parameters that may be used to adjust segmentation results.
Color Threshold - A smaller value will result in more segmented regions.
Luminance Weight - Importance of luminance with respect to chrominaces.
Minimum Region Size - Regions smaller than this size will be merged with their neighbors.
- Define Object : To define your desired objects by clicking or drawing box on the segmented image. Note that segmented regions will be shown in random colors.
- Get Object : To get an image with the selected object. Following are a few examples.
If you have any sugestion or encounter any problem, please send emails to dzhong@ee.columbia.edu.
