Impact of Feature Selection Algorithms on Blind Image Quality Assessment

Abstract

Blind image quality assessment (BIQA) is a challenging task in real-world problems due to unavailability of reference images. The performance of BIQA techniques is highly dependent on features used to assess the image quality. In the literature, different BIQA techniques have been proposed using a two-step approach, i.e., feature extraction in different domains and prediction of quality score using extracted features. However, optimum feature selection for these techniques has not been explored. This paper investigates the impact of feature selection algorithms on the performance of BIQA techniques. In contrast to existing techniques, the proposed methodology follows a three-step approach. Firstly, features are extracted using existing BIQA techniques. In the second step, feature selection algorithm is applied on the extracted features to reduce the number of features. The selected features are then utilized for prediction of a quality score in the third step. The proposed approach is evaluated for six BIQA techniques using five commonly used feature selection algorithms. Experimental results show that the feature selection algorithms not only reduces the number of features but also improves the performance of the stateof- the-art BIQA techniques.