<p>The development of a highly intelligent and efficient strategy for image indexing and retrieval from a large image class has become a significant research topic due to the proliferation of digital images and internet technology. In this paper, limitations in image indexing and retrieval from a large image class are addressed using an Autoencoder Apportion-Hash approach and Amalgam index with Key-term Weight Ranking technique. The Micro-Structure Descriptors (MSD) of the annotated training images are used to train the Autoencoder to transform the high-dimensional MSD feature into a low-dimensional feature vector. The hash function for encoding the low-dimensional feature vectors to their respective hash-codes is learned using an Apportion-Hash structure. Using the Autoencoder Apportion-Hash approach, the images are indexed into the relevant hash bins. To retrieve the images using a text query, the Amalgam index structure is designed to search for key-terms of the text query. The images linked with the key-terms are retrieved and ranked based on their key-term weights. By using performance metrics like Mean Average Precision, Precision, Recall, and F-measure, the proposed strategy has been compared with other methods that were previously in practice. The strategy proposed performs superior to existing approaches based on experimental data.</p>