论文部分内容阅读
针对传统视觉注意模型在遥感影像视觉显著区域检测中存在的计算复杂度高、检测精度低等缺点,提出了一种新的视觉显著区域快速检测算法。首先利用整数小波变换降低遥感影像的空间分辨率,从而降低视觉注意焦点检测的计算复杂度;然后在视觉特征融合中引入二维离散矩变换,生成边缘与纹理信息更为丰富的遥感影像显著图;最后在显著图分析中提出区域增长策略来获得视觉显著区域的精确轮廓。实验结果表明,新算法不仅有效降低了遥感影像视觉显著区域检测的计算复杂度,而且能够精确描述视觉显著区域的轮廓信息,同时避免了对整幅遥感影像的分割与特征提取,为今后的遥感影像目标检测提供了一定地参考价值。
Aiming at the shortcomings of traditional visual attention model such as high computational complexity and low detection accuracy in the detection of visual significant regions in remote sensing images, a new fast detection algorithm of visual salient regions is proposed. Firstly, integer wavelet transform is used to reduce the spatial resolution of remote sensing image, so as to reduce the computational complexity of visual attention detection. Secondly, the two-dimensional discrete moment transform is introduced into the visual feature fusion to generate a significant image of the remote sensing image with richer edge and texture information Finally, a regional growth strategy is proposed in the saliency map to get the exact contour of the visual salient region. The experimental results show that the new algorithm not only effectively reduces the computation complexity of the detection of the visual significant region of the remote sensing image, but also can accurately describe the contour information of the visual salient region and avoids the segmentation and feature extraction of the whole remote sensing image for future remote sensing Image target detection provides a certain reference value.