To perceive shape from shading correctly, information about the direction of illumination is needed. We investigated how the direction from visual information affects perceiving shape from shading and how we estimate the direction of illumination using both visual information and the Light-from-above assumption. The results showed that the information about the direction of illumination was extracted from the surrounding visual stimuli and that the area within, where this information propagated, was limited. In addition, even when the explicit visual information about the direction of illumination was available, the Light-from-above assumption also worked, at least in some situations.The effect of assumption was enhanced by the surrounding stimuli with luminance gradation. These results indicate that the configurations of visual stimuli largely affect the estimation of the direction of illumination for shape from shading.
An object-based watermarking scheme is described. Watermarks are embedded into objects by employing object features : a center of an area and outline. Our schemes are available for the MPEG-4 standard to access objects directly within a video sequence. They are also available for one of the second generation watermarking schemes to improve robustness by employing significant parts or features of the data in the watermarking process. Experimental results show the methods are robust against object geometric manipulation such as scaling, rotation, and shape subsampling and are also robust against a Stir Mark 3.1 attack.
A secure printing technique is described that can identify illegally copied or tampered printed documents. Two kinds of micro vector patterns are faintly printed all over the document as a watermark, where it vanishes upon duplicating the document. The watermark can identify whether it is an original or not and, moreover, check whether it has been tampered with or not because the watermark contains the timestamp and original document information. Our experimental results show that the micro vector patterns are simple and effective detectors and can verify the logitimacy of documents in a short rotation.