Emotion Semantic Annotation Of Cheongsam Images Based On Multi-features

Human understanding for an image is based on the semantic level,and with a certain degree of subjectivity.Emotional semanteme is the highest level of semantic meaning in image semantics,which occupies an important position in the research of image annotation.Because of the semantic gap,the mapping between image features and sentiment semantics is extremely difficult,which restricts the realization of emotional semantic annotations.Therefore,emotional feature extraction and semantic mapping become the key technologies and research difficulties in the field of image emotional semantic annotation.This thesis analyzes the technology of image emotion semantic annotation based on its application status.As the uniqueness of cheongsam image,a method of emotion semantic annotation based on Multi-features is proposed in this paper.With the combination of the brightness,the saturation,the warm and the cold membership and the color contrast,we obtain the color features.The color features and texture feature are integrated into the emotion features of the image through weight combination.And then,the image emotion space is established through cognitive psychology and factor analysis experiments.We use PSO-LSSVM for regression build the relationship between feature space and emotional space of the image.The experimental results show that the method described in this paper can significantly improve the accuracy of image emotional semantic annotation.The sentiment semantic of cheongsam image can be effectively identified,and the semantic gap between the image and emotion can be narrowed.