Regional watermarking in spatial domain

Histogram specification based marking considers as watermarks positive integer functions defined on the graylevel range. Image watermarking consists in exact histogram specification. Robustness against attacks is assured when watermarks are almost zero on compact intervals, i.e., groups of graylevels are very little represented in the marked image. Two types of such watermarks are considered: sawtooth shaped and histogram with notches. This paper approaches histogram based watermarking in a regional context: instead of the entire image, only a region is marked. The histogram of the other region is exactly specified in order to preserve image original histogram. The marked region is selected to be simply connex and it is assumed to contain essential details of the image. Two strategies for region extraction are discussed: the inner quarter of the image or a true region surrounded by edges. While the former is computationally very simple border effects do appears at marking, effects which are completely avoided by the latter strategy. Besides the statistical invisibility of the method, the results obtained so far shows robustness.