基于改进Canny算法的女性服装款式轮廓检测研究

Research on Edge Detection of Women's Garment Style Based on Improved Canny Algorithms

为了降低女性服装多样性(如纹理、款式等)在边缘提取环节带来的难度,提出一种改进Canny算法,使得服装款式图像的轮廓提取更具有实用性和强适用性.创建女装图像的样本库后,对样本进行图像预处理,并使用Otsu算法对所获得的灰度图像进行二值化,凸显出目标的服装轮廓.然后在Canny算法中采用自适应的双阈值进行边缘分割,提升边缘的准确性,并设计追踪算法,提取服装款式的外部轮廓,减少了纹理和服装所形成的噪声.根据实验结果,对比微分算子算法和改进Canny算法.该算法在女性服装款式轮廓提取时,表现出良好的鲁棒性、准确性和连续性.

In order to reduce the difficulty in edge extraction brought about by the diversity of women's wear(e.g.,texture,style),this paper proposes an improved Canny algorithm that promises to make the contour extraction of clothing images more practical and applicable.Firstly,a sample library of women's wear images is created,and the samples therein are preprocessed.After that,the grayscale images obtained are binarized with Otsu algorithm with the aim of highlighting the contour of the targeted clothing.Then,this paper conducts edge segmentation with the improved Canny algorithm,designs a tracking algorithm and extracts the outer contour of clothing in different styles.Experimental results show that compared with differential operator algorithm,the improved Canny algorithm shows better robustness,accuracy and continuity in the contour extraction of women's wear.