基于加权K近邻算法的抽象画图像情感分布预测

现有的大部分图像情感分类计算主要致力于预测图像情感的类别,没有考虑观察者对于图像情感不同的主观感受,因此仅对图像进行情感分类并不能满足现实需要.本文提出采用加权K 近邻算法对数据集中每幅抽象画图像进行离散情感的分布预测,首先提取图像的情感特征,不同的K 值,按照距离加权为每幅图像预测对应的情感分布情况,然后与数据集已知的情感分布进行比较.以Abstract 图像库作为数据集进行实验,并验证了算法的有效性.

Emotion Distribution Prediction of Abstract Painting Image Based on Weighted KNN Algorithm

Most existing works on image emotion analysis mainly focused on assigning a dominant emotion category to an image. However, there is often insufficient, because the observer's perception of the image emotion is very subjective and different, only the emotional classification of the image can't meet the needs of the reality. In this paper, we use weighted KNN algorithm to predict the discrete emotion distribution of each image in the data set. Firstly, we extract emotion features of each image, different K values are weighted according to distance to predict the corresponding emotional distribution of each image and compare it with the known emotion distribution of the data set. Abstract image database is used as the data set to test and verify the effectiveness of the algorithm.