Transcript Image Database Access

Image Database Access
Find images from personal collections
Find images on the web
Find images from medical cases
Find images from art collections
Find images from architectural cases
Stockman MSU CSE
1
General methods of query
Use prestored symbolic keys – standard
Use example images provided by user
User specifies colors, textures, shapes
User specifies image regions
User specifies region relationships
User sketches structures of images
Stockman MSU CSE
2
Query by example
 User provides
image (top
left)
 System
creates its
own feature
rep. to match
to other
images
Stockman MSU CSE
3
QBIC (IBM) color histogram
matching; user chooses colors
Stockman MSU CSE
4
Query is grid painted by user
Stockman MSU CSE
5
Texture features also possible
Stockman MSU CSE
6
User can sketch objects (more
research needed)
 User sketches
boundaries of interest
 System will use
elastic matching (see
Ch 14 S&S) on
images in DB
 Can be expensive
Stockman MSU CSE
7
Results of elastic matching
Stockman MSU CSE
8
Current problems
 Indexing needed for fast browsing, but how can
indexes be built?
 Computing image features online will be slow,
even offline computing will be slow.
 What about deeper queries: “show me pictures of
children enjoying eating” (same problem faced
by traditional databases)
 Show me pictures of tragic events, of poverty, of
natural beauty, of triumph against bad odds …
Stockman MSU CSE
9