Feature Selection for Automatic Image Annotation

Abstract

Abstract. Automatic image annotation empowers the user to search an image database using keywords, which is often a more practical option than a query-by-example approach. In this work, we present a novel image annotation scheme which is fast and effective and scales well to a large number of keywords. We first provide a feature weighting scheme suitable for image annotation, and then an annotation model based on the one-class support vector machine. We show that the system works well even with a small number of visual features. We perform experiments using the Corel Image Collection and compare the results with a wellestablished image annotation system.

Dynamic Adaboost learning with feature selection based on parallel genetic algorithm for image annotation

“…Lu et al use GA for feature selection based on MPEG-7 standard, the accuracies of image annotation system are improved significantly [10]. Setia et al use a Gaussian mixture model to weight the features effectively [13].…”

“…In order to get rid of redundant information and improve the performance of image classification, feature selection techniques are used in many researches [10,12,13,23]. Lu et al use GA for feature selection based on MPEG-7 standard, the accuracies of image annotation system are improved significantly [10].…”

“…Feature selection is to find the most representative feature subset or weighted feature subset for image classification [10,12,13]. Traditional feature selection is to select an optimal subset of features.…”

Feature selection based-on genetic algorithm for image annotation

“…Setia et al formatted image annotation with keywords as a classification problem and used a Gaussian mixture model to weight the features effectively [9]. In [10], Qi et al applied likelihood normalization to optimize weights for Corel images automatically.…”

Conservation of effort in feature selection for image annotation

“…In contrast other approaches have used machine learning techniques to identify features or tune a classifier based on performance often using a wrapper approach [7] during the classification/training phase. Setia and Burkhardt [8] focus more on the configuration, tuning and weighting of a feature subset based on a quantitative computed measure of likelihood that describes the similarity of a feature and its discriminative ability. This is implemented using a wrapper approach with a support vector machine.…”