基于分形纹理特征和小波变换的网状纹理检测方法

Netlike Texture Detection Method Using Fractal Texture Features and Wavelet Transform

网状纹理广泛存在于日常生活中,其独特的网格结构以及极易发生形变的特点使网状纹理的描述和检测成为一件困难的工作．本文以球网检测为例,提出了基于分形纹理特征和小波变换的网状纹理检测方法．先对足球视频图像进行小波多分辨率分解,计算不同尺度图像的分形纹理特征向量,基于因果关系对特征进行选择组成特征向量．然后使用支持向量机检测图像中的网状纹理．实验表明此方法有较强的鲁棒性,能够在剧烈形变、复杂背景和基元大小差异明显的条件下成功检测出球网．

Netlike texture is very common in our daily life. However, it is very difficult to be described and detected because of its distinct characteristics. This paper chooses nets in sports game as research object and presents a netlike texture detection method of using fractal texture features and wavelet transform. First, multi-scale wavelet decomposition is employed on video frames; then fractal features of multi-resolution images are obtained. A group of fractal features are chosen as feature vector according to causality theory and an SVM classifier is used to recognize the net in images. The experiments show that this method is robust and effective under the condition of intense distortion and complicated background. This method can be extended to other netlike texture detections.