Image Quality Assessing by using NN and SVM

In the correlative curve of image subjective and objective quality assessing, there are some points that lower the performance of image quality assessing model. In this paper, the concept of isolated points was given and isolated points predicting was also illuminated. A new model was given based on NN-neural network and SVM-support vector machines with PSNR and SSIM-structure similarity, which were used as two indexes describing image quality. NN was used to obtain the mapping functions between objective quality assessing indexes and subjective quality assessing value. SVM was used to classify the images into different types. Then the images were accessed by using different mapping functions. The number of isolated points was reduced in the correlative curve of the new model. The results from simulation experiment showed the model was effective. The monotony of the model is 6.94% higher than PSNR and RMSE-root mean square error is 35.90% higher than PSNR