基于视觉底层特征的图像增强方法

Image Enhancement Based on Low-level Visual Features in the Visual Region of Interest

目的提出基于视觉底层特征对不同类型图像权重的图像增强方法。方法提取图像的视觉底层特征,如颜色、亮度、方向、纹理和边缘特征;加权融合成计算显著图;与眼动仪测得的眼动显著图进行相关系数比较,以确定各视觉底层特征的最佳权重。再根据权重的大小为不同类型的图像选择合适的图像增强方法。结果采用基于视觉底层特征的图像增强方法,其增强后的效果更符合人眼视觉感知。结论针对不同类型的图像需要充分考虑各视觉底层特征的权重大小,使其能真实反映视觉感兴趣区域,以达到提高图像增强后的视觉感知一致性。

Objective To propose an image enhancement method based on the low-level visual features according to their weights in the different types of image. Methods Firstly the low-level visual features of the image were extracted,such as color, intensity, orientation, texture and edge features, and then computing saliency map was produced by weighted integration. The optimal weights of the low-level visual features were then determined by comparison with eye tracker measured human saliency map using the correlation coefficient. Finally, the appropriate image enhancement methods were selected based on the weights for different types of images, in order to improve the consistency of visual perception.Results Experimental results showed that the image enhancement method based on low-level visual features used in this paper had a more consistent enhanced effect with human visual perception. Conclusion We need to fully take into account the feature of the low-level visual features for different types of images, so that it can truly reflect the visual area of interest in order to improve the consistency of visual perception after image enhancement.