When you search for images, you can compare them based on any combination of their properties: color, shape, texture, and object. For example, if you had a collection of handbag images, you could retrieve all the images that contain red bags, bags of a particular shape, plaid-patterned bags of a certain shape, etc.
In the eVe system, images have four distinct attributes: color, shape, texture, and object region. Each of these attributes is used in the corresponding search type. Note that size does not factor into the attributes. Images in eVe are scale-independent to ensure more accurate comparisons.
The eVe engine automatically segments an image into distinct object regions, then generates scale-independent descriptions of those regions, known as visual signatures. An images visual signature is a double array vector composed of the calculated values for its four basic properties of color, shape, texture, and object.
These visual signatures are organized in a proprietary indexing scheme for extremely fast retrievals. A comparison of images is done by comparing the visual signatures.
To generate a visual signature for an image using the eVe API:
1. Prepare images for analysis (see the Additional Configuration section in the Installation Guide).
2. Instantiate a MediaObject.
3. Load an image into the MediaObject.
4. Pass it to the analyze () method within the analyze class.
See the Analyzing a File section
for an example of how to analyze an image. See the eVe
SDK API Reference Guide for more information about the analyze
() method.
When you perform a search, you can specify the relative importance of each of the four basic properties. You can also specify the order in which you want eVe to return the results of your search. For example, if you were searching for pictures of a particular car, you could weight the shape and region properties heavily, and perhaps not include color at all.