<p>Image-in-image hiding (IIH) embeds a secret image into a target image via reversible data hiding, generating a camouflage image with high visual quality while enabling lossless reconstruction of the secret image. Existing image transform-based IIH methods typically sort image blocks in ascending order of standard deviation and perform sequential matching. However, standard deviation cannot accurately characterize the similarity of pixel values between blocks, which often leads to considerable matching errors. This paper proposes an IIH scheme based on locally optimal block matching. Specifically, the target image blocks are first clustered according to their variances, which in turn guides the clustering of secret image blocks. Within each cluster, a greedy strategy based on local search is adopted to select the optimal matching block by minimizing the mean squared error, thereby improving matching accuracy. Experimental results demonstrate that the proposed method effectively enhances matching accuracy and reduces the storage overhead of auxiliary information, while maintaining high visual quality and reversibility. Experiments on the UCID, DIV2K, and COCO datasets demonstrated that this method achieved average PSNR values of 22.10 dB, 21.99 dB, and 20.64 dB, surpassing existing state-of-the-art methods by 0.16 dB, 0.2 dB, and 0.17 dB, respectively.</p>