Research On Stereo Image Quality Assessment Method By Binocular Fusion Of Human Visual Perception

With the development of society, the image plays a more and more important role in people’s life. Especially, people have become accustomed to new visual experience such as3 D movie and 3D game with the time, how to meet the needs of people on the stereoscopic image quality has become an important topic in the field of image processing. Therefore, the quality assessment method of stereoscopic images has become a pressing issue. In order to get a better result, it is very important to us to study effective evaluation method. Here are some exploration and results about the research on the quality assessment of stereoscopic image, the main work are as follows:By analyzing the physiological and psychological characteristics of the human eye,human visual system and the principle of contourlet transform, we proposed a new blind stereoscopic image quality assessment method based on the disparity map and contourlet transform. Firstly, this method introduces two methods of extracting disparity map:subtraction and the global error energy minimization algorithm. then, we extract three features from distorted image and disparity map by contourlet transform. The features are energy,energy differences between scales and measurements of the statistical relationships of pixels within scales. Finally, we use support vector machine(SVM) to establish the image quality evaluation model and get the quality of image. Experimental results show that this method can achieve more satisfactory results than classical methods and related methods in recent years. And what’s more, this method has better subjective consistency.No-reference stereoscopic image quality assessment method based on the disparity map and complex contourlet transform is proposed. Firstly, we introduced into the disparity map that was extracted by the second method, then decomposed image by the complex contourlet transform and extracted the following features: the energy and the energy differences between scales. Finally, those features are fed to SVM to predict image quality. Experimental results show that the proposed method outperform current 3D IQA methods and the first method we proposed.We propose a no-reference stereoscopic image quality assessment method based on the saliency map and disparity map. Firstly, we introduced into the disparity map, then we extract energy, energy differences between scales from distorted image, disparity map and saliency map by the complex contourlet transform. Finally, we get quality of image by the model that is accomplished by SVM. Experimental results show that the proposed method has better subjective consistency than current 3D IQA methods and the former two methods.