Image Emotional Semantic Classification Based on Line Direction Histogram
Jianchao Zhang · Jisuanji gongcheng · 2005
Image semantic classification using low-level features is an important problem in semantic-based image retrieval. Although semantic description and classification in emotional way have become remarkable in recent years, the study in this field is still at the very beginning. This paper shows how high-level emotional representation of art paintings can be inferred from perceptual level features suited for the particular classes (dynamic vs. static classification). According to the strong relationship between notable lines of image and human sensations, the edge-based line direction histogram is selected as the image feature, and probabilistic neural network (PNN) is used to establish the mapping between image feature and semantic description, then images can be classified into dynamic vs. static. Experimental results demonstrate the effectiveness of the approach.