Image Fusion Based On Wavelet Transform Digital Watermarking Technology

As the power of computers increases rapidly, computers are used widely. As the by-product of fast Internet expansion, the security of personal information is jeopardized. Conventional data encryption techniques can partially prevent unauthorized release of information. However, the progress in deciphering techniques put almost all data encryption techniques on the edge of crisis. There are a huge number of hackers who cannot help to decode any encrypted data brought to their attention. Encrypted data face the same risk as plain data. Moreover, conventional encryption techniques cannot follow and check the validity and copyright when the carrier data are used in gear. Currently, lots of researchers use algorithms analogous with animals' disguising colors to hide information. The risks for the hidden information are reduced greatly. Firstly, except the author himself, others will not notice that there is information hidden in the carrier data. Therefore, there is no incentive for them to decode the hidden information. Secondly, the hidden information will not propagate even though the carrier data are damaged. As a result, a new field, that is, information hiding, is explored.Digital watermarking is one of available techniques to hide information. It embeds some data into the carrier data. The embedded data can be checked at any time with certain methods. It requires that the embedded information will not interfere with the normal use of the carrier data. This feature is called invisibility. At the same time, the carrier data may be compressed to transmit, and processed to satisfy various needs. The watermarking information should still be able to be checked. The feature is called robustness.Image fusion technology attracts researches' attention due to large embedding volume, easily adjustable embedding intensity, good invisibility, etc. Using Equation (1), one image can be hidden into another image. When necessary, the hidden image can be extracted by using Equation (2). In above equations, I is the original image; J is the hidden image, I' is the fused image; and t is fusion factor. The fusion intensity can be adjusted by changing the value oft.However, many algorithms were proposed in the spatial domain. When fused in the spatial domain, the hidden image is vulnerable to image attacks, In order to avoid degrading the quality of the carrier image, the fusion intensity is usually reduced, which reduces the algorithm's robustness as well. Currently, most fusion algorithms determine the fusion factor through trial and error. The fusion factor determined in such a way is relatively small. As a result, the algorithms' capability of resisting attacks is often sacrificed.In order to improve the hidden image's capability of resisting attacks, this study analyzes fusion models and improves conventional fusion algorithms. Because wavelet transform leads to good time and frequency characteristics for signal and image processing applications, this study investigates image fusion algorithms in frequency domain. To fuse images in wavelet domain, the first problem to solve is how to tell whether the images lose fidelity or not. Concepts such as peak signal to noise ratio are not applicable any more in wavelet domain. The second problem to consider is how to increase the images' fusion factor in wavelet domain and improve the hidden image's capability of resisting attacks without losing image fidelity.,.Through a large amount of experiments, following conclusion is drawn: when peak signal to noise ratio in the spatial domain is PSNR=28, the mean square error (MSE) in wavelet domain is the corresponding threshold. Based on above fact, the equation to calculate the fusion factor in wavelet domain is derived.One practical algorithm is proposed. It is called segment fusion algorithm. The algorithm is based on wavelet transform.Segment fusion algorithm is proposed to exploit the fact that, after images are wavelet transformed, most energy concentrates at the low frequency end. Firstly, the carrier image and the hidden image are wavelet transformed once. Then, the two images' low frequency portion is divided into segments of the same size. The average of coefficients is calculated for each segment. The hidden image's low frequency segments are moved and sorted according to the relative magnitude of segment coefficient average in the carrier image whose low frequency segments are not moved. Consequently, the two images' fusion factor is improved and the fusion intensity is increased.Numerical experiments show that Segment fusion algorithm increase the images' fusion factor. The hidden image's capability of resisting attacks is improved without affecting the carrier image's fidelity. The algorithms' robustness is improved as well. The algorithms can be used widely in copyright protection for electronic image products and digital libraries.