In content-based image retrieval (CBIR) applications, each database needs its corresponding parameter setting for feature extraction. However, most of the CBIR systems perform indexing by a set of fixed and pre-specific parameters. On the other hand, feature selection methods have currently gained considerable popularity to reduce semantic gap. In this regard, this paper is devoted to present a hybrid approach to reduce the semantic gap between low level visual features and high level semantics, through simultaneous feature adaptation and feature selection. In the proposed approach, a hybrid meta-heuristic swarm intelligence-based search technique, called mixed gravitational search algorithm (MGSA), is employed. Some feature extraction parameters (i.e. the parameters of a 6-tap parameterized orthogonal mother wavelet in texture features and quantization levels in color histogram) are optimized to reach a maximum precision of the CBIR systems. Meanwhile, feature subset selection is done for the same purpose. A comparative experimental study with the conventional CBIR system is reported on a database of 1000 images. The obtained results confirm the effectiveness of the proposed adaptive indexing method in the field of CBIR.