一种改进型Canny联合阈值分割的图像背景去除——在外观专利图像检索中的应用

Improved background removal based on Canny algorithm and threshold segmentation

背景去除是外观专利图像检索中图像预处理的重要归一化步骤,针对外观设计专利图像库中图像背景多样性、复杂性和随机性的特点,提出了一种改进型Canny算子联合阈值分割的图像分割算法。通过对图像颜色采样的自适应阈值分割获取阈值,联合Canny算子和形态学方法对图像生成前景物体分割模板,从而有效去除背景信息。实验证明,与改进前的图像分割算法对比,该算法得到的前景物体轮廓更加精确;针对外观专利图像检索平台中的图像分割,该算法具有很好的自适应性。

In the design pattern image retrieval system, background removal is the important step of image normalization when the image is reprocessed. An improved algorithm of Canny operator combined with threshold segmentation is introduced, considering the properties of image in the design pattern image database, that is variety, complexity and randomicity. Through an adaptive threshold segmentation based on sampling with color to acquire the threshold, it combines with Canny operator and morphological method to get the foreground object segmentation template. The experiment shows that the algorithm introduced in this paper generates more accurate contour of foreground objects in contrast with the algorithm not being improved. For image segmentation of design pattern image retrieval specially, this algorithm shows good adaptability.