Robustness and Imperceptibility Enhancement in Watermarked Images by Color Transformation

Maedeh Jamali, Mahnoosh Bagheri, Nader Karimi, Shadrokh Samavi Isfahan University of Technology Isfahan, 84156-83111 Iran

Abstract—One of the effective methods for the preservation of copyright ownership of digital media is watermarking. Different watermarking techniques try to set a tradeoff between robustness and transparency of the process. In this research work, we have used color space conversion and frequency transform to achieve high robustness and transparency. Due to the distribution of image information in the RGB domain, we use YUV color space, which concentrates the visual information in the Y channel. Embedding of the watermark is performed in the DCT coefficients of the specific wavelet subbands. Experimental results show high transparency and robustness of the proposed method.
Keywords—watermark, digital media, robustness, color transform.
I. INTRODUCTION
Watermarking is a protection technique that is used to decrease the concern of digital data copyright, such as video, audio, and images. Image watermarking means hide an image into another image. They named the watermark image and host image, respectively.
Watermarking methods can be classified into several categories. One of them is blind or non-blind. That means whether or not additional information is needed for watermarking. The proposed method in [1], falls into a non- blind category because for extract watermark image from a watermarked image, the host image is needed. But in [2,3] the methods are blind and can extract watermark images without the need for the host image.
The other category is the domain that the watermark image uses for the embedding process. There are two domains: spatial and transform. Spatial domain methods are usually less susceptible to conventional image processing attacks because the information is directly embedded in image pixels and easily manipulated. There are various techniques in the transform domain such as Discrete Cosine Transform (DCT) [3]-[5], Fourier Discrete Transform (DFT), Discrete Wavelet Transform (DWT) [6]-[8], Contourlet Transform (CT) and Hadamard Conversion [2].
In [3], authors introduce an adaptive blind watermarking in which the watermark is embedded in DCT coefficients of CT. Two-level CT is applied to the host image. They divided the approximate image into blocks in first levels. Then they extracting the important edges of each block using their proposed edge detection method. These areas are candidate regions for strong embedding. Some parts of the second level are also concatenating with the mentioned blocks. The entropy of blocks and some other image metrics of each block produce an adaptive strength factor for that block. Finally, the DCT transform of blocks is used in their embedding algorithm. In [4], Fang et al. find the relationship between positions and the magnitude of changes in the DCT coefficient and direction of tissue blocks. The direction factor mapping designed by examining such a relationship. The method proposed in [6] is a non-blind watermarking. It uses the HL sub-band of the first DWT of 512×512 image to embed a watermark image with a size 32×32. A geometric algorithm is proposed for embedding to generate a trade-off between robustness and imperceptibility in [8]. They used eight samples of wavelet approximation coefficients from each image block and built two line-segments in a two-dimensional space. Some methods only work in one domain, and others work in a combination of domains. Some approaches use multi transforms. For example in [9], DCT and DWT transforms are used together. The proposed method is a combination of DCT, DWT and fuzzy system for embedding. They first transformed the image into DWT for two levels and selected the LL region in both transformations. Then divide a selected area into 8×8 blocks and calculate the DCT of them. Three attributes related to HSV are fed to a fuzzy system to calculate the strength factor for each DCT block.
After that, they manipulate the DCT coefficients based on these adaptive strength factors and watermark bits. Some papers use color image watermarking space conversion to increase robustness. There are several color spaces like RGB, HSI, Lab, YCbCr. Usually, authors prefer to use YCbCr color space because the Human Visual System (HVS) is more sensitive to luminance than color. In RGB color space, all colors have the same resolution, but in YCbCr, Y has high resolution, but Cb and Cr have lower.
In [5], the host and cover images are color images. The authors convert the color watermark image to binary image, and for the host image, convert it to YCbCr and use channel Y for embedding. The channel Y is divided into 8×8 blocks and after apply DCT to each block

…(Full text truncated)…