基于面料图像梯度的服装风格相似性匹配算法

针对不同服装风格图像的相似性匹配问题,传统方法停留在定性分析和主观评价阶段,无法定量、客观地判定服装风格。为了对服装风格相似度进行定量分析,提出了一种基于面料图像梯度的服装风格相似性匹配算法。该算法首先基于面料灰度图像对图像的梯度进行计算。然后设置阀值对图像梯度进行过滤处理,得到属于不同风格类型主要特征的图像梯度。接着按X方向和Y方向进行分区对主要特征情况进行统计,获得每个区块的梯度波动空间频率,并计算出整体图像的空间频率方差。最后选择以图像X、Y方向梯度方差的百分比为权重得到图像风格的整体相似度评价指标。实验证明该算法利用面料图像对服装风格相似性分析识别具有一定可行性。

In view of the similarity matching of images of different clothing styles,the traditional method stays in the stage of qualitative analysis and subjective evaluation,and it is impossible to judge the clothing style quantitatively and objectively.In order to quantitatively analyze the similarity of clothing style,a matching algorithm for clothing style similarity based on the gradient of fabric image is proposed.The algorithm f irst calculates the gradient of the image based on the grayscale image of the fabric.Then set the threshold to f ilter the image gradient,and get the image gradient which belongs to the main features of different style types.Then the main features are calculated according to the X-way and Y-direction partition,the gradient f luctuation space frequency of each block is obtained,and the spatial frequency variance of the overall image is calculated.Finally,the overall similarity evaluation index of image style is obtained by weighting the percentage of image X and Y-direction gradient variance.Experiments show that the algorithm uses fabric images to analyze and identify clothing style similarity.