Research Of Image Quality Assessment For Underwater Transmission Based On Reduced-Reference And No-Reference Image Quality Assessment

With the development of science and technology, human activities have expanded from land to ocean which occupies the70%of the world, therefore the study of underwater acoustic communication technology becomes more and more urgent. In the study of underwater acoustic communication studies, images and other multimedia services has become a hot topic in underwater acoustic communication. Due to the uncertainty of the ocean water environment, scattering in the water and absorption effects of nonlinear effects, it will result in the image quality distorted, especially easily to cause the image blurred. In addition to this, multipath channel distortion in underwater communication will also cause the image distorted in the transmission process. All above this, they will affect the display on the video terminal. So it is significant to develop a appropriate method of image quality evaluation for the entire underwater imaging technology in the underwater channel.Firstly, this paper sum up the conventional image quality assessment algorithms, including subjective evaluation methods and objective evaluation methods, and classify the previous objective evaluation methods, and make a conclusion that a good evaluation algorithm is consistent with the subjective evaluation scores. Apart from this, this paper make an introduction about the human visual system, the popular subjective image quality assessment databases, the popular learning method basing on the support vector machine. Secondly, according to the characteristic features of the image before and after embedding into the original image which reflect the deterioration in the original image and the human visual system of the location personality, this paper proposes a reduced reference image quality evaluation method based on significant region of the image. The proposed method is simple, and it only needs to transfer the small amount of data. Finally, this paper summarizes the causes of the blurred images and analyzes the fuzzy type. And according to the fuzzy characteristics, this paper uses the values of the GLCM in the gradient image to reflect the level of blur. After this, this paper utilizes the SVM to train the data, and raising the no-reference fuzzy image quality evaluation method based on the supported vector machine learning.On the basis of the above methods and research, this paper adopts the popular international mainstream LIVE2, CSIQ, TID2008databases to validate the proposed methods and use the Pearson, Spearman indicators to verify the performance the proposed method. It proves that the methods proposed have good subjective similarity. Apart from this, using the methods, this paper builds the related GUI platform so as to predict any deterioration image.