### 1. Introduction

Region of interest (ROI) and region of non-interest (RONI) watermarking: In the ROI watermarking techniques, the watermark is embedded in the ROI in such a way that the perceptual quality of the image is not compromised. The watermark information is embedded in the RONI in order to keep the ROI distortion free. This way the diagnosis value of the medical image is not compromised [2,3]. In medical images the RONI generally contains the black background that encircles the ROI.

Reversible watermarking: The second approach corresponds to reversible watermarking. Once the embedded content is read, the watermark can be removed from the image, which allows for the retrieval of the original image [4].

Classic watermarking: The third approach consists of using classical watermarking methods while minimizing the distortion. In this case, the watermark replaces some image details, such as the least significant bit of the image [5,6], or some details are lost after lossy image compression [7].

### 2. Visual Cryptography and FS-DWT Transform

### 2.1 Visual Cryptography

### 2.2 Image Watermarking Based on Visual Cryptography

### 2.3 FSDWT Transform

### 3. Mammograms Preprocessing

### 4. Proposed Watermarking Approach

### 4.1 Used Techniques

#### 4.1.1 Dominant blocks

*σ*′≥

*Tσ*, where T is a parameter,

*σ*′ and

*σ*, respectively, the standard deviation of the transform coefficients and the local deviation for a given 7×7 block.

#### 4.1.2 Reversible Walsh-Hadamard transform

**SS**and an embedding map

**EM**containing the coordinates of the dominant blocks. To merge these two matrixes in a reversible way, we used the reversible Walsh-Hadamard transform of order N=2, which leads to a reconstruction without distortion using the following forward and inverse transform matrices [19]:

*x*,

*y*) be the constructed pairs from

**EM**and b the binary number from

**SS**:

### 4.2 Concealing Process

**Inputs**: Original Image

**I**(m×n), Watermark Image

**W**(r×c), Secret Key

**S**

**Outputs**: Private Matrix

**PM**(r, 2×c)

**Step 1:**Perform the FSDWT transform on the image**I**, and find all the dominant blocks.**Step 2:**Use S as a seed to select random r×c dominant blocks.**Step 3:**Construct an embedding map**EM**, such that the entries in the matrix are the positions of the selected dominant blocks obtained in the above step.**Step 4:**Construct a feature image**F**, such that the entries in the matrix are the sample averages of the selected dominant blocks. Let F_{avg}be the average of**F**.-
**Step 5:**Construct a binary matrix B: **Step 6:**Use the bits in matrix**B**to select columns in Table 2 for generating the secret share**SS (r,2xc).****Step 7:**Merge the matrix**EM**with the matrix**SS**using the Reversible Walsh-Hadamard Transform.**Step 8:**Use S as a seed to construct a noised private matrix**PM**from the matrix obtained in the above step.

### 4.3 Extracting Process

**Inputs**: Attacked Image

**I’**(m×n), Secret Key

**S**, Private Matrix

**PM**(r, 2×c)

**Outputs**: Watermark Image

**W’**(r,2×c)

**Step 1:**Perform the FSDWT transform on the image**I’**.**Step 2:**Use**S**as a seed to denoise the private matrix**PM**and extract the embedding map**EM**and the secret share**SS**.**Step 3:**Construct a feature image**F**, such that the entries in the matrix are the sample averages of the dominant blocks. Let F_{avg}be the average of**F**.-
**Step 4:**Construct a binary matrix B: **Step 5:**Use the bits in matrix**B**to select columns in Table 2 for generating a public share**PS**. Note that the code-block assignment for a public share corresponding to each secret bit is independent of the pixel pair colors in the watermark image.**Step 6**: Perform logical OR on the public share**PS**and the secret share**SS**to extract the watermark.

### 4.4 Reduction Process

### 5. Experimental Results

### 5.1 Robustness Test

_{i,j}and c′

_{i,j}denote the pixel intensity of the original and attacked images.