图像分辨率对天然纤维识别的影响

Image Resolution Influence on Nature Fiber Recognition

针对低分辨率图像的保真问题,提出基于轮廓曲线的邻域方向差分链码,进行曲线失真程度的分析.该方法将基于像素坐标的二维轮廓曲线转化成相对于曲线走向而变化角度函数.差分链码具有平移、拉伸和旋转不变性.根据差分链码提出分别针对于对象轮廓周长、轮廓包围面积和轮廓曲线相似性的误差统计模型,并将该模型应用于棉麻等纤维截面的显微图像在不同分辨率下外形的失真度分析,给出在图像采集时允许的最低图像分辨率的计算.

A novel approach, the differential neighborhood chain, is presented for evaluating the distortion of the outline of the nature fiber in the low resolution image. The 2-D edge curve is represented by a function that describes the curve direction with the curve length. The representation has three characteristics： i. e. , rotation invariant, scaling invariant, and translation invariant. With the differential neighborhood chain, the circumference error, the area error, and the similarity are proposed. The error models are applied to the natural fibers and the error is studied. The minimum image resolution calculation for the natural fiber is discussed.