With the growing interest in cross-modal retrieval technology, cross-modal hashing has become a mainstream trend for comparing and searching between different modalities. However, when faced with multi-label information, existing research has often neglected the unique features of each individual label, resulting in learned hash codes lacking discriminative power. In this article, we present a novel cross-modal hashing method called Supervised Discriminative Discrete Hashing (SDH ), which consists of two stages: hash code learning and hash function learning. Specifically, we extract each individual label from the multi-label information and introduce their one-to-one corresponding hash code. By mapping the original multi-label information to Hamming space through the single-label hash codes, we can obtain more discriminative to-be-learned hash codes. Furthermore, we establish a connection between label information and original feature vectors to explore the potential consistency. Additionally, we use a discrete optimization algorithm to reduce information loss and quantization error during the learning process. The superiority of our proposed SDH method over several state-of-the-art hashing methods has been demonstrated through extensive experiments on two benchmark datasets.