The article describes enhancements in retrieval performance of content-based image retrieval (CBIR) system using the fuzzy class membership-based retrieval (CMR) framework. The CMR approach explores the CBIR as a classifier-based retrieval problem using a neural network classifier, accompanied by a simple distance-based retrieval method. The fuzzy class membership-based approach is known to enhance the retrieval performance along with slight variation without any constraint on the feature set to be used. Despite that, its efficacy is not known for color and multi-band textures. We have proposed several advancements in a fuzzy class membership-based retrieval framework for improved retrieval. The main contributions are the simplification of vital threshold selection process and effective use of membership values to encourage the use of appropriate classifiers, investigation of the role of the cost function in neural network and distance weighting functions for improved retrieval, a way to adapt a new classifier in fuzzy class membership-based retrieval framework in place of neural network. Experimental analysis of all proposed advancements are evaluated using benchmark gray-scale texture databases viz. three versions of Broadtz and Outex database. The p-value analysis is carried out to check if the improvements are statistically significant. The proposed method is further tested with the Describable texture database (DTD) and Multi-band texture (MBT) database to check its applicability on color textures. The comparison with recent methods using gray-scale image databases viz. AT&T face database, MIT VisTex database, Broadatz texture database, and natural-color image databases viz. Corel-1K and Corel-10K showcase the efficacy of the proposed method.