Quality Assessment Of Stereoscopic Image Based On GA And SVM

In recent years, with the rapid development of stereoscopic image technology,3Dmovies and virtual reality technology has been integrated into people’s daily work,study and life. The stereoscopic image quality assessment is one of the key issues ofthe stereoscopic image technology.Because the amount of data of3D image is larger than2D image, the thesis firstlyuses dimensionality reduction through principal component analysis, which leads toeigenvectors reflecting the essential characteristics of the stereoscopic image. That isthe principal component of stereoscopic image feature space. Secondly, usingBT-SVM model based on statistical learning theory classifies the quality level of thestereoscopic image. The performance of BT-SVM system is affected by its structure,so this article determines the structure of BT-SVM by separately measure samples infeature space. In addition, the classification result of the model is largely affected byparameters, for the SVM model parameters are set mainly based on the prioriknowledge. Therefore, in each SVM subcategories training process of the BT-SVMsystem, this experiment uses the genetic algorithms to choose, crossover and domutation operations to the parameters, screening out the optimal parameters of themodel, and finally, get the excellent BT-SVM model.The simulation results show that, the classification accuracy ratio of the modeloptimized by GA can reach94%. The proposed scheme can offer better classificationeffect of the objective assessment of the stereoscopic image, and the effect is inaccord with the human eye subjective feelings more.