To perform a visual search on an image from a search result:
1. After selecting a category, select the image attribute upon which you want to search. The four attributes upon which you can search include color, shape, texture, and object:
A color search is based upon the visual similarity of the predominant colors within images.
A shape search is based on the two-dimensional outlines of objects within an image that are used as patterns to match against other images.
A texture search is based on unique textures within an image that is used to rank other images based on their similarity to that texture.
Within an object search, eVe groups statistically similar parts of an image into a certain number of regions, or classes. The classes in the source image are then compared against the classes within other images to determine visual similarity.
Note: In the sample application, you can select one attribute upon which to search only. When using the full eVe SDK to add visual search functionality to an application, you can enable users to define a combination of attributes upon which to search.
2. Select a threshold for the search from the Threshold drop-down list. The value you select for the threshold determines what degree of visual similarity that the search should use as criteria when retrieving/comparing images. A high threshold means images must be very similar, thus fewer images are retrieved from the search; a low threshold relaxes the criteria, thus more images are retrieved from the search.
3. Click Zero In under an image in the Search Results section to initiate the search. The results of the search are displayed.