The Research Of Underwater Image Restoration

The technology of digital image restoration has achieved remarkable results in many areas, whereas, research on the underwater image restoration is very rare, which is due to the complexity of environment under water. Having entered the twenty-first century, people come to realize the importance of role played by ocean for a country to rise. The requirements for digital image processing have become higher and higher with the development of areas like marine geographic engineering survey, marine monitoring, marine environment detection and underwater vehicle. Due to the attenuation and scattering effect occurred in the acquisition process, the quality of the original image would get degenerated seriously. Thus, how to restore the original image accurately and how to obtain useful information has become an important research project. This thesis is elaborated as follows:To begin with, the thesis introduces the optical transmission characteristics under water as well as the influence they bring. In addition, the primary feature of underwater imaging is analyzed using a qualitative approach.Secondly, A brief summary on the basic theory of underwater image restoration and the point spread function is given, with enumerating several typical point spread function with the help of quantitative approach. Next, this paper briefly introduces the concrete steps of the classic image restoration algorithm and underwater image quality assessment standards.Finally, This thesis put forwards three ways to restore images under water and a comparison is made among these three different methods. The first method is based on the backscatter. By building a physical model of how images degrade under water, which is to divide the target subject and the receiving sensor in the water into n layers, and after which, add the additive noise and underwater images that goes through n-tie point spread function using calculus, we could make an underwater restoration of black and white images under water using the Wiener Filtering under the circumstances that both the point spread function and distance under water are unknown. And this restoration is taken on two conditions, with one being far from the object, and the other near the object. The second method uses a model, which the first method proposed. But this time, the spread function and underwater distance is not unknown any more. And we make a deconvolution recovery using the Wiener Filtering and the Least Square Filtering. The last method is to increase regularization constraint items and airspace adaptive factors, which is based on original NAS-RIF cost function in order to achieve a image restoration under water. Finally, the three methods are compared to identify the most suitable method for underwater image restoration.