Watermark Hiding in HDR image based on Visual Saliency and Tucker decomposition

Abstract Recently, great attention has been paid to high dynamic range (HDR) images because of their richly detailed and high dynamic range of intensity. The need for pre-processing of the HDR image format with tone mapping (TM) operators makes it unique; the TM operators are provided on display with a low dynamic range (LDR). On the other hand, TM can be regarded as an inevitable attack when protecting HDR image ownership is considered. An adaptive approach for concealing watermarks based on visual saliency and Tucker decomposition has been presented in this article. In the first step, three feature maps were produced and split into three color components (RGB) using the host image. Considering that out of the three feature maps, the first one represents the most significant part of the information in the HDR image, it was selected due to its robustness in terms of watermark insertion. Afterward, an HDR image saliency is created to identify the most significant locations that can be used to securely embed data to achieve a balance between visual quality and watermark strength. The salience map is produced by decomposing a specific HDR image into three LDR images with various intensity levels using a tone-mapping operator. Consequently, the contrast feature (CF) model is applied to each decomposed LDR image; then, the three saliency maps are integrated into a single map. Results of the experiments show that the proposed technique demonstrated superior performance compared with the extant algorithms for HDR image watermarking in terms of resisting various attacks while considering invisibility and embedding capacity