In recent years, the intersection of deep learning, hashing, and retrieval systems has witnessed significant advancements, particularly in the realm of image processing and information retrieval. In this paper, we present a novel end-to-end trainable network using contrastive masked autoencoder (CMAE) for efficient image retrieval. We comprehensively investigate the integration of contrastive masked autoencoders with hashing techniques, coupled with various types of losses, namely HashNet loss, Deep Supervised Hash (DSH) loss, and Greedy Hash loss, for enhancing retrieval performance. We delve into the efficacy of different masking methods in conjunction with these techniques to facilitate efficient representation learning. We investigate HashNet loss, a novel objective function tailored for learning hash functions directly from data, and contrastive loss, which encourages similar items to have similar hash codes while pushing dissimilar items apart. First, we introduce the concept of masked autoencoders, a variant of traditional autoencoders designed to learn robust representations from partially observed input data. We explore various masking strategies, such as attention masking, random masking, and patch masking, elucidating their effects on the encoding process and subsequent retrieval performance. Furthermore, we present a comparative analysis of different retrieval methods, including cosine similarity, knn approach and content-based retrieval approach, within the context of contrastive masked autoencoder method on different benchmark datasets like CIFAR10, ImageNet, MS-COCO and NUS-WIDE in terms of retrieval accuracy (mAP). The code is available at https://github.com/Mehulk43/CMAEH .