基于LBP和小波纹理特征的室内室外场景分类算法

Indoor-outdoor Scene Classification Algorithm Based on the Texture Feature of LBP and Wavelet

室内室外场景识别是图像处理的基本问题之一。对此问题提出了一种不需要图像分割和目标识别的方法。首先,利用小波变换对原图像进行五层小波分解,然后用旋转不变的LBP(local binary pattern)算法对第二层至第五层的LL子图提取纹理特征,再计算第二层和第一层LH,HL,HH子图的能量均值和方差,最后连接这两个过程形成的低维特征向量进行室内室外图像场景分类。实验结果表明此算法分类效果比较好,且识别室内室外场景并不需要目标形状等局部详细信息,只需要图像模糊化的全局纹理信息。

The scene recognition between indoor and outdoor is one of the basic problems in image processing. A solution to this problem which does not require image segmentation and object recognition is proposed in this paper. First of all, a 5-level wavelet transform was implemented on the image, then the rotation invariant LBP algorithm was used to extract the texture feature from 2-level to 5-level LL sub-images. Afterwards, the mean and variance of the LH, HL, HH sub-images from 2-level to l-level were computed. Finally, these two low-dimensional eigenvectors were concatenated as the image feature to classify the indoor and outdoor scene images. The experiment results indicate that our algorithm performs well on indoor/outdoor classification, and the overall texture information of the fuzzy image is only needed in the scene recognition between indoor and outdoor, rather than the local detailed information such as object shape.