视觉显著性检测的研究

Research on Visual Saliency Detection

计算机模拟人类视觉系统的生理机理对图像进行识别是视觉显著性检测研究的主要目的之一。在对图像的预处理过程中,快速准确的显著性检测能够为图像检索、图像识别、图像分割等后期图像处理过程提供有价值的参考。然而目前视觉显著性检测的算法还存在检测结果质量不高的问题。通过研究图像傅里叶频谱与显著性的关系,提出一种基于相位谱和调谐幅度谱结合分水岭算法的显著性检测方法,以提高目前视觉显著性检测的效果。

It is one of the main purposes of visual saliency detection to simulate the physiological mechanism of human visual system. As the preprocessing of image, reliable and rapid saliency detection can provide valuable reference information for image retrieval, image segmentation, image retrieval and so on. However, at present, there are some problems in the algorithm of visual saliency detection. Through the study of the relationship between Fourier spectrum and the significance of the image spectrum, proposes a phase spectrum and amplitude spectrum tuning saliency detection method based on watershed algorithm, to imorove the visual saliencv detection effect.