Introduction
- Digital Image watermarking is a technique to hide the secret image (watermark image) into cover image resulting watermarked image.
- Watermark image has to sustain against various attacks on watermarked image.
Steganography is about concealing their very existence.
Watermarking, as opposed to steganography, has the (additional) requirement of robustness against possible attacks
Watermark
- Watermark image can be imprinted on the cover image either by visible or invisible means
- Watermark image may be Binary, Gray or color image
- Size of the watermark image that can be embedded depends on the algorithm used for watermarking
Visible
- An image (watermark) that is overlaid on the primary (cover) image, which is visible in the watermarked image
Invisible
- An image (watermark) which is overlaid on the primary (cover) image, which is invisible but which can be detected algorithmically
Logo or seal of the organization which holds the rights to the cover image, it allows the cover image to be viewed, but still marks it clearly as the property of the owning organization.
Invisible Watermark- Embedding level is too small to notice
- Can be retrieved by extraction algorithm
- Applications:
- Authentication
- Copyright Protection
- Broadcast Monitoring
- Copy Protection
- Imperceptibility
- Robustness
- Inseparability
- Security
- Capacity
Types of Watermark in terms of Fidelity
- Robust watermark
- Fragile watermark
- Semi-fragile watermark
- This watermark has the ability to withstand to various attacks on watermarked image thus providing copyright protection.
- A robust watermark should be stuck to the document it has been embedded in, in such a way that any image transform of reasonable strength cannot remove the watermark. Hence a pirate willing to remove the watermark will not succeed unless they debase the document too much to be of commercial interest
- This watermark is mainly used for detecting modified data of the watermarked image.
- This watermark gets degraded even for a slight modification of data in the watermarked image.
- Useful for authentication
- It is an intermediate between fragile and robust watermarks. It is not robust against all possible image attacks.
- Spatial domain
- Transform domain (DFT, DCT,DHT,SVD,CT,DWT)
- Hybrid domain(Combination)
Simple Technique: LSB substitution Technique
- To obtain the bit planes of the cover Image
- Replace the zeroth bit plane of cover image with watermark image
Advantages in Spatial Domain Technique
- Requires no watermark image to retrieve it from watermarked image
- No blocking artifacts
- Advantage: Maximum Capacity
- Prone to tampering and attacks on watermarked image like
- compression
- rotation
- scaling
- translation
- cropping etc.
- Cover image is transformed into a another domain using DCT, FFT, SVD,CT and Hartley etc.
- Watermark image is embedded in the transform coefficients of the cover image
- Watermarked image is obtained by taking the inverse transform after embedding watermark image in transform coefficients
- Watermark is extracted from the watermarked image by taking transform and extracting the watermark image from the coefficients
- Robustness
- Resistant to rotation, scaling and translation
- Resistant to JPEG Compression
- Imperceptibility
- Less Capacity
- Computationally Complex
- Blocking artifacts due to block processing
Hybrid domain technique
- Advantages of both spatial and transform
- Flexibility in embedding
- Combine the characteristics of different transforms.
- Non-oblivious(Non-blind): Both the cover image and secret keys used in the extraction.
- Semi-oblivious(Semi-blind): The secret key(s) and the watermark image or some side information used in the extraction.
- Oblivious (Blind): Only the secret key(s) used in the extraction.
- Canny’s edge detection algorithm is applied to the entire cover image of size NxN. Size of the watermark image is MXM.
- The edge detected cover image is divided into nxn non over-lapping blocks.
- Number of edges in each block is computed. Blocks are arranged based on descending order of the number of edges in each block.
- The first M2 blocks having more number of edges are selected and indexed for watermark embedding.
- SVD transformation is applied on each selected individual block.
- The watermark image is permuted with a secret key K
- A matrix of is formed with largest singular values of each selected block.The entire range to is divided into various bins as shown in Table 3.1. Here, and indicate the minimum and maximum values of respectively. A step size of is taken as the difference from one bin to another bin.
- Each element of matrix is checked for its position in Table 1.
- After identifying the bin number, is modified as follows:
- After the modification applied to the largest singular values, inverse SVD is applied to unmodified U, modified D, and unmodified V, to get the watermarked image.
- Extraction Algorithm
- The watermarked image of size NxN is partitioned into blocks of size nxn.
- Blocks having highest number of edges are identified in the watermarked image.
- SVD transformation is applied on each selected block
- A matrix of Dlarge is formed with largest singular value of each selected block.
- Quantization table as shown in Table 1 is generated at the receiver.
- From the bin position obtained in Step 5, each Dlarge value is Checked for its position in the quantization table (Table 1). If it is in Range 1 ,the watermark bit is ‘1’, otherwise it is ‘0’.
- Stirmark
- Optimark
- Checkmark 2.1
- Extension to video watermarking
- Improving the imperceptibility
- Improving robustness to image attacks like printing and scanning
- Faster algorithms
- Improved quality assessment metrics for watermarked images
No comments:
Post a Comment