A Blind Digital Watermarking Method for Omni-directional Images using Feature Points

摘要

Assuming that the owner of each original image may be different, inserting different watermark data into each original image, and extracting a watermark from the projected image in which the omnidirectional image is serviced, A blind digital watermarking method, comprising: (a) obtaining a plurality of original images for an omnidirectional image; (b) selecting an image portion to be watermarked by selecting a portion of each original image that is not used for stitching; (c) extracting feature points from the selected image portion; (d) setting a peripheral area of the feature point as a DW target area; (e) selecting a position to insert an actual WM in the DW target area; (f) inserting a watermark at a predetermined insertion position; And (g) inversely transforming the image resulting from the insertion of the watermark. By using the watermarking method as described above, by using feature points of regions not used for stitching and using data around each feature point as a DW position, it is possible to have robustness against image distortion generated in the omnidirectional image generation process. Sufficient robustness for JPEG compression attacks.