eVision logo
Visual Search Technology Unleashing the power of visual e-business
News Technology & Products Business Solutions Developers About eVision

Also See:

Visual Search Whitepaper
(1.2 MB pdf)

General FAQ

Developer FAQ

Online Documentation


eVe Demos

Obtaining good search results

Future Directions

 home > technology > overview > partial image search example
Technology Overview - Partial Image Searches

With eVe 3 Professional a user can now select specific regions in the image to more accurately control the visual search. When eVe analyzes an image, it automatically segments it into object regions. If the user does not tell it otherwise, eVe will search for other images that have visually similar characteristics across all object regions. If however the user selects specific regions, eVe will search for images that have visually similar object regions. This selective search is called partial image search.

Below is an example of how partial image search can narrow down your query. The first set of images looks at the entire image and finds all other images in the database that are similar in color in the object regions. (Note: because of the object nature of eVe, it still only finds images that are faces, even though the database contains a wide variety of people, places, objects and beach photographs).

In the second set of images, we specifically want to search for woman with black hair in strapless dresses, so we select the strapless dress, and the skintone. The results that eVe returns are images of woman in in dresses with bare shoulders.

eVe Object-based Search using color similarity
(all object regions included)
Query Image
object map
Object Map
girl07 face9 face34 face32
Top 4 search results ordered by similarity

eVe Partial Image Search using color similarity
(skin and dress object regions selected - shown here by horizontal lines)
query image
Query Image
object map with selected regions
Object Map
girl07 girl03 girl06 face14
Top 4 search results ordered by similarity