A Watermarking Algorithm in Wavelet Domain Based on Texture Measures

Abstract

Joint Noticeable Difference (JND) threshold is usually used to control the embedding location and intensity in image watermarking system, and to determine the extraction location in blind algorithm as well. The robustness is increased if the watermark is embedded in areas where texture is complex. Texture measures are done on the low frequency coefficients of wavelet domain to determine the areas where the watermark is to be embedded. A novel JND computation method based on the low frequency coefficients of wavelet domain and texture measures is presented to enhance the robustness of JND threshold. This algorithm enhances the invisibility of watermark and obtains an optimum value of PSNR. Most of the embedded watermarks can be effectively located even when the image is largely altered, thus it can improve effectively the detection result in blind extraction. Besides, a novel blind detection algorithm based on the sign of values of diagonal detail subband coefficients is proposed.This detection enhances detecting responses and obtains high credibility.