实数型曲线描述子二值化研究

Binaryzation of real-valued curve descriptor

曲线匹配技术在模式识别、计算机视觉和图像理解中具有重要作用。随着移动设备的广泛使用,有必要研究存储空间小、匹配速度快的二值型曲线描述子。针对常见实数型曲线描述子(MSCD、IOMSD、IOCD和TCHP),利用阈值化方法,获得由0、1表示的二值曲线描述子。实验结果表明,在旋转、视角变化和光照变化条件下,提出的曲线二值描述子能够保持实数型描述子的匹配准确性,而占用的内存空间仅为原描述子的1/32或1/16。

Curve matching plays an important role in pattern recognition, computer vision and image understanding. With the widespread use of mobile devices, it is necessary to study the binary curve descriptors with small storage space and fast matching speed. We quantize the curve descriptors, such as MSCD, IOMSD, IOCD and TCHP, into binary strings represented by 0 and 1 by thresholding the real- valued vectors of the descriptors. Experimental results show that under the conditions of image rotation, viewpoint change and illumination change, the proposed binary descriptors yield a little higher matching accuracy while only requiring 1/32 or 1/16 storage space of the original curve descriptors.