In this work, an attempt has been made to differentiate human epithelial type-2 (HEp-2) specimen images using Bag of Features (BoF) and Adaptive Cuckoo Search (ACS) feature selection. For this, 420 images consisting of homogenous and speckled patterns were obtained from a publicly available International Conference on Pattern Recognition (ICPR) 2016 database. These images are preprocessed using edge-aware local contrast enhancement and subjected to a speededup robust feature (SURF) descriptor for feature extraction. The optimal features are identified using the ACS method and are then fed into a support vector machine (SVM) for classification. The results show that the proposed approach is able to distinguish homogenous and speckled patterns. It is found that the features identified using ACS-based feature selection are significant. The proposed approach yields an average accuracy of 97.90% using the SVM classifier. Because automated analysis and classification of HEp-2 specimen images is important for the diagnosis of autoimmune diseases, this study seems to be clinically relevant.