Transillumination imaging is commonly used in medicine and biometrics to provide non-invasive insights into internal structures. However, the prevalent image blurring resulting from scattering effects poses a significant challenge to the effective application of transillumination. Artificial intelligence algorithms have gained prominence for enhancing transillumination images and addressing challenges such as scattering suppression, depth estimation, and three-dimensional structure reconstruction. These advances require high-quality input images to optimize model performance. Acquiring a large-scale transillumination image dataset for practical AI applications is difficult due to subjective imaging conditions. This study aimed to overcome this obstacle by introducing a comprehensive dataset of transillumination images. Methods and algorithms for generating depth-dependent point-spread function and transillumination images were presented. The dataset comprised 12,000 pairs of images of clear and scattered media, each associated with the corresponding depth information. This study is valuable for advancing AI-based solutions in transillumination imaging and provides a foundation for further research on image deblurring, depth perception, and 3D reconstruction.