CAJ | 학술논문

本文在传统词袋模型的基础上，结合人的视觉特性，提出了一种基于视觉显著度与词袋模型的图像分类方法。算法首先计算图像的视觉显著度，然后根据图像的视觉显著度对图像计算视觉单词的加权直方图，然后使用视觉单词的加权直方图表示图像。通过在Caltech 101数据库进行实验，验证了本文方法的有效性，实验结果表明，该方法能够大幅度提高图像分类的性能。
본문재전통사대모형적기출상，결합인적시각특성，제출료일충기우시각현저도여사대모형적도상분류방법。산법수선계산도상적시각현저도，연후근거도상적시각현저도대도상계산시각단사적가권직방도，연후사용시각단사적가권직방도표시도상。통과재Caltech 101수거고진행실험，험증료본문방법적유효성，실험결과표명，해방법능구대폭도제고도상분류적성능。
Based on Bag of Words model,an improved algorithm based saliency is proposed in this paper. Firstly, computing saliency image of the original image, according to the saliencyimage, we weighted visual word histogram of the image, and then use the weighted histogram visual to represent the image. The experiments arecarried out on Caltech 101 database. The results show that the proposed method performsbetter than the traditional method.