BP神经网络在图像语义自动标注的应用

Image Semantic Annotation Based on BP Neural Network Algorithm

由于传统的基于内容图像检索存在的"语义鸿沟"问题,其在某些特定的领域无法满足用户的需求。图像语义自动标注的出现能够有效地解决这方面的问题。该文提出了先使用Normalized Cuts方法对图像进行区域分割并提取出每个区域的低层视觉特征,再利用BP神经网络算法来学习图像区域和标注字的对应关系来进行图像语义的自动标注的方法,实验结果证明了此方法的有效性和准确性。

Because of the "Semantic gab" problem in the traditional CBIR(content-based image retrieval) systems,they are not fit in some special areas.The appearance of Image-Semantic-Annotation can effectively resolve this problem.This paper represents a method that first uses the Normalized Cuts to cut the image into several areas and then extract the low-level features,last use BP neural network algorithm to learn the relations between the areas and the annotation words.The experiment's results proved this method is valid and efficient.