A Survey of Thumbnail Image Reconstruction Techniques: A Forensic Perspective

Thumbnail images are small previews generated by operating systems, mobile gallery applications, image viewers, cameras, and image-processing software. In digital forensics, thumbnails are important as they remain available even when the original image is deleted, moved, overwritten, or stored on external media no longer attached to the device. In image reconstruction research, thumbnails represent a difficult low-resolution input: they are usually down-sampled, cropped, compressed, and stripped of much of the high-frequency content required for faithful high-resolution recovery. This work reviews the literature on thumbnail image reconstruction techniques ranging from direct recovery and classical interpolation to CNN-based super-resolution, GAN/perceptual methods, super-resolution, transformer-based restoration, diffusion-based super-resolution. The review presented in this paper includes identifying a research gap with respect to metadata-aware, uncertainty-aware, and hallucination-aware thumbnail reconstruction systems designed to focus various forensic review.