Large scale image tamper detection and restoration

Sarkar, Dipabali; Palit, Sarbani; Som, Sukalyan; Dey, Kashinath

doi:10.1007/s11042-020-08669-0

Cited by 13 publications

(1 citation statement)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These methods not only determine if an image has been tampered with but also identify where the tampering occurred and can pinpoint the specific areas. Other researchers [22][23][24][25][26][27][28], have suggested watermarking methods that enable the identification of tampered areas and image recovery. However, the methods used to embed the watermarks can also damage the original image.…”

Section: Introductionmentioning

confidence: 99%

Reversible Medical Image Watermarking for ROI-Based Tamper Detection and Recovery

Arham¹,

Nugroho

Adji

2023

Preprint

View full text Add to dashboard Cite

Medical image watermarking has become a widely studied topic in recent years. The method used to embed data into medical images must be reversible becouse tampered in medical images caused by irreversible embedding can lead to inaccurate diagnosis. Medical image watermark- ing should be able to identify the tampered area and recover it. This research work introduced a method for reversible watermarking of med- ical images. This method can embed data, detect any tampering that occurs, locate the area that has been tampered with, and then recov- ery that area of tampered in the medical image. The medical image is segmented into three sections. The first is the region of interest (ROI) for reversible embedding of patient records. The second is the region of non-interest (RONI) for reversible embedding of localization data for tampered and recovery areas. The third is the border area used for embedding coordinates of the selected ROI using Least Significant Bits (LSB). The proposed method has a capacity of up to 6 bpp of data and 12 % of the ROI area with very good visual quality of up to 50.86 dB. The watermarked image, when extracted, is identical from the original image except for the border area, provided that there is no tampering. If the ROI is tampered, the area of tampered can be identified and restored with results reaching up to +27.61 dB, with only a small difference in image quality -2.86 dB compared to the original image after extraction.

show abstract