Toward robust watermarking of scalable video

the maximum correlation value is chosen. The location giving maximum correlation value for the correct source estimates the exact scanning location. Two main issues in this extension of source camera identifi cation methods [3-5] are increase in computational complexity in estimating the correlation between reference pattern and test noise patterns of different sizes and overheads in estimating the row reference patterns for the complete width of the scanner's bed (where fi le sizes can grow to more than 100MB). Use of linear scanning elements in fl atbed scanners makes the whole process feasible as only row reference pattern needs to be estimated and not the estimation of the complete two dimensional reference pattern (covering scanner bed of size 8.5*11.7 square inches). This also reduces the search complexity in testing phase from O(N2) to O(N), when the image covering the complete scanner bed has O(N2) pixels. Experiments done on images scanned at 200dpi from ten different scanners show close to 90% classifi cation accuracy for testing images scanned from different random locations of the scanner bed. Further experiments to extend this scheme for forgery detection are under progress. To detect the forged regions of a image claimed to coming from a particular scanner, fi rst the scanning location (in horizontal direction) is determined by applying above steps and then each row of the noise pattern of the image in consideration is correlated with relevant section of the row reference pattern. The group of rows with comparatively lesser correlation (an experimental threshold can also be determined) is expected to correspond to forged region. We are also working on evaluating the robustness and accuracy of the proposed algorithm with varying scanner resolutions, JPEG compression and on images undergone fi ltering operations like sharpening, contrast stretching and resampling. Artifact-metrics is an automated method of authenticating artifacts based on a measurable intrinsic characteristic. Intrinsic characters such as microscopic random-patterns made during the manufacturing process, are very diffi cult to copy. Since the fi ber distribution of paper is random, a transmitted light image of the distribution can be used for artifact-metrics. Little is known about the individuality of the transmitted light image, although it is an important requirement for artifact-metrics. Measuring individuality requires that the intrinsic characteristic of each artifact signifi cantly differs, so having suffi cient individuality can make an artifact-metric system highly resistant to brute force attack. Here we investigate the infl uence …