Abstract:

Combining user perception and semantic processing on a computer of image resources, is an important and feasible way to narrow the “semantic gap” in it. This paper presents an image annotation method by multidimensional semantic visualization, which visualizes the multidimensional semantics of an image as a point in a Cartesian coordinate system, converts user’s input on the Cartesian coordinate system to image semantics based on nonlinear programming and multi-objective programming, and supports perceptual human-computer interacting annotation of image semantics by this bidirectional multidimensional semantic visualization. Experiment results state the high accuracy, helpfulness and user satisfaction in image annotation, which proves our method’s conduciveness to narrow semantic gap in image annotation.

Key words: Image annotation, 　Image semantics, 　Semantic visualization, 　Semantic gap, 　Human-computer interaction