High-quality recovery of digital images

Abstract Advancements in technology have significantly simplified the manipulation and modification of digital images. In the last decade, a lot of research has been done to counter image tampering. One technique that has been used successfully is digital watermarking. The research in detecting tampering using watermarking has significantly matured; however, high-quality recovery of sections of the original image that have been tampered with still remains a challenge. Watermarking images for tamper detection is also a very time-consuming process, especially in the context of live streaming. In this paper, a quick watermarking-based scheme is proposed that can not only detect tampering in an image but can also recover the tampered portion with a high peak signal-to-noise ratio (PSNR). The proposed scheme was tested on some commonly used images in image processing and a number of datasets. The results show that the watermarked images had a PSNR of well above 37dB for all eight datasets even when the third least significant bits were used. We also compared our model with the state-of-the-art and existing fragile self-recovery methods. Results show that the proposed method could recover images with a PSNR of 38.36dB, while earlier methods were only able to achieve a PSNR $\leq $ 36dB, indicating a significant improvement in image quality. Finally, the proposed method is quick and results indicate that the proposed method is three to six times faster in detecting and recovering images than the base studies. Thus, the proposed scheme represents a notable advancement in image forensics and tampered image recovery.