Abstract: This paper proposes concepts of unitary projection histogram and unitary projection histogram moment invariant, and constructs a unitary projection histogram moment invariant for image feature extraction. Experimental results show that compared with the classic moment invariant theory, it can reveal structural information more effectively and distinguish subtle differences and similar images better. And it is robust when applied in image recognition.

Key words: unitary projection histogram, unitary projection histogram moment invariant, image recognition, moment invariant, Hu moment invariant

摘要： 提出归一化投影直方图和在平移缩放变换下保持不变的归一化投影直方图不变矩概念，由此构造归一化投影直方图不变矩，用于提取图像特征。实验结果证明，与经典不变矩理论相比，该方法能有效反映图像的结构特征，识别细微差异和区分相似图形的能力更强，应用于图像识别时具有较好的鲁棒性。

关键词: 归一化投影直方图, 归一化投影直方图不变矩, 图像识别, 不变矩, Hu不变矩

CLC Number: 

• TP391