基于侧抑制频谱调谐的显著性检测方法

传统的基于频域的显著性检测方法存在处理结果不稳定以及缺少仿生学意义等问题。根据人眼识别的侧抑制机制，提出了一种频域调谐的显著性检测方法。该方法通过对图像傅里叶频谱进行多种非线性自适应调谐，达到抑制图像中冗余特征及增强显著特征的效果，进而实现图像显著区域的有效检测。通过对自然图像和心理物理学模板图像的对比实验和分析表明，该方法在显著区域检测率、显著区域轮廓完整性及显著区域对比度等方面，都比现有显著性检测方法有较大提高。

Saliency Detection Method Based on Spectrum Tuning Using Lateral Inhibition

The traditional spectral-based saliency detection methods have many drawbacks, such as unstable detection result and lack of bionics meaning. This paper presented a new saliency detection method based on spectrum tuning using lateral inhibition mechanism. The new method tunes the Fourier spectral self-adaptively using multiple nonlinear tune forms to inhibit the redundant image features and enhance the saliency image features, and therefore the salient region could be detected effectively. We tested this method on both real natural images and artificial pictures such as psychological patterns.The results indicate that our method is more efficient than other methods in aspects of saliency region detection rate, salient region contour detection and high salient region contrast.