Content-based retrieval of paper defect images
- The goal is to develop tools for management and content-based retrieval of paper defect images in huge image databases.
- Inspection systems may produce thousands of images in a day.
- Analysis is impossible without efficient databases and clustering methods.
- Using PicSOM, similar defect images are searched from a database based on the similarity of their visual features
- Features may measure different visual properties of images, e.g. their color, texture, or edgeness
- No manual annotation nor labeling of images is needed
Examples of Defect Classes
Examples of PicSOM Queries
