基于第二代小波变换的医学图像检索

Medical Image Retrieval Based on Second-generation Wavelet Transform

纹理特征提取是基于内容的图像检索的一个重要环节.针对该问题,提出了一种简单而有效的纹理特征提取算法.采用原始图像的亮度以及图像经过第二代小波变换后子带的方差信息来描述纹理.与传统的采用第一代小波变换进行图像检索的算法相比,提取的特征维数少,计算复杂度小.通过对医学图像库的检索实验结果表明,该算法具有高效的检索效率.同时,还具有对图像的平移、旋转、尺度以及镜面变换的近似不变性.

Texture feature extraction is an important part in content-based image retrieval. This paper proposes a simple and effective texture feature extraction algorithm. It uses the luminance property of the original image and the variance informantion of different subbands of the second-generation wavelet transform to form the descriptor. Compared with the previous retrieval algorithm using traditional wavelet transform, this algorithm has low-dimensional feature indexing and little computation. Experimental results on the medical database show that the proposed algorithm is greatly effective in image＇s retrieval. Furthermore, it is almost invariant to translation, rotation, scale and mirror.