一种改进的图像重复模式检测方法研究

重复模式是城市景观的突出显著特征,重复模式的检测对于城市景观分析具有重要意义。提出了一种通过综合使用特征提取及直积低秩建模实现图像重复模式检测的新方法。首先,利用消失点和Harris角落,从正面图像中自动选择出具有代表性的纹理实现图像的矫正,然后提出基于直积的新算法,可以检测出正面图像纹理内的重复模式。最后,基于大量图像对方法进行了测试。实验结果表明,方法可以显著提升检测质量。

Research on An Improved Detection Method of Image Repeated Patterns

Repeated patterns are prominent and significant features in urban scenes. Detection of these repeated patterns becomes very important for city scene analysis. In this paper, a new detection method of repeated patterns is proposed by combining feature extraction followed by a Kronecker product low rank modeling approach. Firstly, the representative texture within facade images are automatically selected using vanishing points and Harris corners for rectifying the images, secondly, a novel algorithm based on Kronecker product for detection of repeated patterns within that texture of the facade image. Finally, we have tested the proposed method in a large set of images, the experimental results show that, this method can significantly improve the detection quality.