Honeycomb structure artifacts in fiber bundle endoscope images (FB images) degrade image quality and affect the accuracy of clinical diagnosis. Existing algorithms such as spatial domain and frequency domain processing, as well as deep learning-based algorithms, suffer from the problems of reduced image clarity and loss of details after restoration, which make it difficult to fully retain the original information. In this paper, a depixelation and enhancement algorithm of fiber bundle images based on diffusion model is proposed, which can effectively remove honeycomb structure artifacts, while better retaining the texture information of the original image. The experimental results show the proposed algorithm outperforms the existing methods in both PSNR and SSIM metrics, especially the PSNR metrics are improved by 95.4%, 93.4% and 58.9% on the kidney, liver and tonsil tissue datasets, respectively, which verifies its validity and superiority in FB image restoration tasks.