Abstract: In the context of 3D reconstruction of finless porpoises, challenges such as color distortion in underwater images, the limited availability of finless porpoise datasets, and the difficulty in obtaining multi view images of finless porpoises. However, emerging methods have not yet been applied to finless porpoises. To address these challenges, we proposes a single view 3D model reconstruction method for finless porpoises that combines diffusion models and neural radiation fields. Firstly, an improved underwater image enhancement method has been developed, effectively addressing the issue of color cast and distortion in underwater images. Secondly, we created a custom multi view image dataset of finless porpoises and fine-tuned a perspective conditional diffusion model to achieve the synthesis of multi view images from a single view, providing a new approach for reconstructing finless porpoises from a single image. Finally, a 3D model of the finless porpoise was reconstructed using neural radiation field. The results of 3D reconstruction of finless porpoises were compared and evaluated using the average Chamfer distance and normal vector consistency. Our method achieved a lower average Chamfer distance and higher normal vector consistency compared to existing methods, indicating that it more accurately reconstructed the body color and morphology of finless porpoises. Moreover, the underwater image enhancement led to further reductions in average Chamfer distance and improvements in normal vector consistency.