Content-based image retrieval using hybrid k-means moth flame optimization algorithm

Content-based image retrieval plays a key role in many domains and the volume of the image databases increases tremendously; it is very difficult to compare query image feature with every image in the dataset during the retrieval phase. Hence, search space and computational complexity increase which degrades the performance of recognition accuracy. This system investigates various search space reduction techniques, which partition or classify the image collection into a subset of related images. This study proposes an image clustering using the hybrid K-means moth flame optimization algorithm (KMFO). It enhances the performance of the K-means algorithm by assigning the optimum number of clusters and cluster centroids using the number of flames and flame values obtained in MFO. It uses color moments, HSV color histogram, color correlogram, GLCM, wavelet transform, dominant color, and region-based descriptors as feature vectors. The experiments are tested on Corel 1K dataset, and it shows competent results when compared with other retrieval techniques.