Understanding a scene provided by Very High Resolution (VHR) satellite imagery has become a more and more challenging problem. In this paper, we propose a new method for scene classification based on different pre-trained Deep Features Learning Models (DFLMs). DFLMs are applied simultaneously to extract deep features from the VHR image scene, and then different basic operators are applied for features combination extracted with different pre-trained Convolutional Neural Networks (CNN) models. We conduct experiments on the public UC Merced benchmark dataset, which contains 21 different areal categories with sub-meter resolution. Experimental results demonstrate the effectiveness of the proposed method, as compared to several state-of-the-art methods.
Image denoising always is one of important research topics in the image processing field. In this paper, fast discrete
curvelet transform (FDCT) and undecimated wavelet transform (UDWT) are proposed for image denoising. A noisy
image is first denoised by FDCT and UDWT separately. The whole image space is then divided into edge region and
non-edge regions. After that, wavelet transform is performed on the images denoised by FDCT and UDWT respectively.
Finally, the resultant image is fused through using both of edge region wavelet cofficients of the image denoised by
FDCT and non-edge region wavelet cofficients of the image denoised by UDWT. The proposed method is validated
through numerical experiments conducted on standard test images. The experimental results show that the proposed
algorithm outperforms wavelet-based and curvelet-based image denoising methods and preserve linear features well.