基于视觉注意的医学图像感兴趣区域提取

Extracting regions of interest in medical images based on visual attention mechanism

提出一种基于视觉注意机制的医学图像感兴趣区域提取方法。受生物学启发,该方法模仿人类自下而上的视觉选择性注意过程,通过计算图像中每个像素的全局对比度,构造相应的显著图,并根据显著图,检测出图像中的显著区域。利用多幅医学图像对本方法进行评估,结果表明,该方法能够快速、精确地提取图像中的感兴趣区域,在提取结果和运算速度上均取得了令人满意的效果。

This paper proposed a new approach for extracting regions of interest in medical images based on visual attention mechanism. Motivated biologically, this approach simulated the bottom-up human visual selective attention mechanism, computed the global contrast of each pixel and constructed the saliency map. According to the saliency map, detected the salient regions in the medical image. This approach had been tested on many medical images. The experiment results show that this approach is efficient both in extraction result and computational speed. Regions of interest in medical images can be extracted rapidly and precisely.