Abstract:

The face image is two dimensional, vertical and horizontal. All objects and object relations are identified prior to their data storages. Each object has been attached with the object number starting from zero (e.g., 0 is face object), and each object exhibits the correlation number starting from zero degree (e.g., 0 degree is the correlation between face and right eyebrow). The number of objects and object correlations is fixed. Object and object correlation exhibit number of attributes (e.g., size at 0 degree, distance between the reference object towards that object).

In this paper a method to handle searching for 2 dimensional face image database (FID) is proposed. FID content will be represented by attributes holding features of objects and object correlation. The objects can be either roundness or non-roundness objects. The proposed method assumes that each image consists of a fixed number of objects and object correlation. The proposed method has several desirable properties:

• With a given cost of matching, a database search is employed by using the proposed method, so that all images with no higher cost will be retrieved.
• The proposed method is much faster than sequential searching for both running either on main memory or on the storage space.
• The proposed method will extremely require less space for storing those objects and object correlation.