Research On Improved Algorithm Of Print Image Emotion Mapping Based On LFCM - SVM

Affective ability is the significant symbol to measure human intelligence, so it is in a long term that human committed to explore how to make computer machine really replace human work. Today, it has been an important issue in artificial intelligence to study and improve the semantic analysis technology of image affective semantic from mechanism of human cognition.Woodcuts as an important category of traditional Chinese painting, they have unique creation skills, beautiful rich colors, and are full of artistic emotion. Thus, the thesis chooses woodcuts as training and testing samples for experiments and puts forward some improvements for affective model, feature extraction algorithm based on color and texture, mapping algorithm based on SVM(support vector machine) from semantic information of image affective semantic.The improvements can be divided into five aspects.1.Filter selection and classification are conducted for experiment samples so as to avoid the influence of judgments based on subjective feelings.2.The thesis actively try to conduct the OHTA color model, to analyze whether the color model is suitable for the improvements of woodcuts image affective semantic or not.3.Dominant color histogram and accumulation color histogram are added in the process of color extraction.4.The thesis takes normalization methods to process gradient and intensity of texture descriptor-the edge histogram.5.LDA(linear discriminant analysis) method is embeded in the preprocess of sample classification. In allusion to above contents, our experiment results focus on the influence that these improvements take to mapping accuracy rate. Actually, compared with former experiment result, the thesis shows the effectiveness of proposed approach, and enhance accuracy rate of the semantic analysis technology of image emotion.