论文部分内容阅读
针对视觉选择性注意模型化计算过程中不同特征在整合阶段的权值判定,提出一种基于特征图分布的权值估计方法,并在静态图像显著性区域提取中取得了令人满意的应用效果.首先提取原始图像的颜色、方向和强度特征图像,然后计算各个特征图的广义高斯分布参数与方差,进而给出一种特征图权值估计算法,最后通过对特征图的加权整合与归一化实现对原始图像的显著性区域提取.实验结果表明,通过此方法计算的权值对特征进行加权调制所提取的显著性区域的效果更加符合人眼的观测结果.
In view of the visual selective attention to the judgment of the weight of different features in the integration stage in the process of modeling calculation, a weight estimation method based on the distribution of feature maps is proposed, and satisfactory results are obtained in the extraction of significant regions of static images Firstly, the color, orientation and intensity feature images of the original image are extracted, then the generalized Gaussian distribution parameters and variance of each feature image are calculated, and then a weight estimation algorithm of the feature image is given. Finally, by weighted integration and normalization To achieve the salient area extraction of the original image.Experimental results show that the saliency regions extracted by weighted weighting of the features by this method are more in line with the human eye observation results.