Single image detail enhancement algorithms based on residual learning often adopt a greedy search strategy, which can easily lead the system to fall into a local optimum. To address this issue, this study proposes a single image detail enhancement algorithm inspired by the foraging behavior of bees. This algorithm models the image block matching process as the behavior of bees searching for high-quality nectar sources in a two-dimensional space. For each image block in the input image, multiple “worker bees” are dispatched to search for the best matching block in the reference image. The matching quality is evaluated by calculating the differences in multi-dimensional features such as context, edge, and texture structure between the candidate block and the original block. By combining the guidance of the global optimal nectar source and excellent individuals in the neighborhood, a search strategy that combines global and local aspects is achieved. The algorithm introduces a probability-driven neighborhood reference selection mechanism and a boundary reflection strategy, and dynamically adjusts the proportion of local and global weights through a cosine function, thereby effectively enhancing the diversity and stability of the search. Experimental results show that after multiple epochs of iterative optimization, the algorithm can extract precise and robust detail features. The proposed algorithm outperforms existing algorithms on multiple public datasets such as RealSRSet, BSDS200, and T91. Specifically, on the RealSRSet dataset with a $\times 4$ magnification, compared to the current popular method QWLS, the proposed algorithm achieves an improvement of 8.21 dB in the PSNR and 0.1569 in the SSIM, demonstrating its advantages in detail restoration and visual quality.