Logo recognition and retrieval algorithm in document image
文本图像中图标图形的识别检索方法

To fulfill the application requirement to retrieve logos in automatic processing document image, a logo recognition and retrieval algorithm based on projective features was presented. The projective histogram was used to denote the horizontal and vertical direction features. To overcome the effects on projective histograms which come from mirror symmetry transform, the similarity was measured by histogram correlation distance in this algorithm. Experimental results show that projective features describe the logo spatial distributing information of horizontal and vertical direction, and can recognize logos effectively.