Joint Dictionary Learning for Multispectral Change Detection

Abstract

Change detection is one of the most important applications of remote sensing technology. It is a challenging task due to the obvious variations in the radiometric value of spectral signature and the limited capability of utilizing spectral information. In this paper, an improved sparse coding method for change detection is proposed. The intuition of the proposed method is that unchanged pixels in different images can be well reconstructed by the joint dictionary, which corresponds to knowledge of unchanged pixels, while changed pixels cannot. First, a query image pair is projected onto the joint dictionary to constitute the knowledge of unchanged pixels. Then reconstruction error is obtained to discriminate between the changed and unchanged pixels in the different images. To select the proper thresholds for determining changed regions, an automatic threshold selection strategy is presented by minimizing the reconstruction errors of the changed pixels. Adequate experiments on multispectral data have been tested, and the experimental results compared with the state-of-the-art methods prove the superiority of the proposed method. Contributions of the proposed method can be summarized as follows: 1) joint dictionary learning is proposed to explore the intrinsic information of different images for change detection. In this case, change detection can be transformed as a sparse representation problem. To the authors' knowledge, few publications utilize joint learning dictionary in change detection; 2) an automatic threshold selection strategy is presented, which minimizes the reconstruction errors of the changed pixels without the prior assumption of the spectral signature. As a result, the threshold value provided by the proposed method can adapt to different data due to the characteristic of joint dictionary learning; and 3) the proposed method makes no prior assumption of the modeling and the handling of the spectral signature, which can be adapted to different data.

Video Salient Object Detection via Fully Convolutional Networks

“…In recent years, the border of saliency detection has been extend to capturing common saliency among related images/videos [40], [41], [42], [44], [47], inferring the salient event with video sequences [39] or scene understanding [48], [49], [43]. However, there are significant differences between above methods and traditional saliency detection, especially considering their goals and core difficulties.…”

Coupled dictionary learning for unsupervised change detection between multimodal remote sensing images

“…This is a challenging task mainly due to the possible dissimilarities in terms of resolution, which prevents any use of classical CD algorithms [44,7], or in term of modality, which makes even flexible CD algorithms inoperative [22,23]. To alleviate this issue, this work proposes to improve and generalize the CD methods introduced by Seichepine et al [43], Gong et al [29], Lu et al [34]. Following the widely admitted forward model described in Section 2.1 and adopting consistent notations, the observed images Y 1 and Y 2 can be related to two latent images…”

“…In the proposed methodology, the CD is based on the reconstruction error of patches approximated thanks to estimated coupled dictionary and independent sparse codes. Following the same principle, in Lu et al [34], a semi-supervised method was used to handle multispectral image based on joint dictionary learning. Both methods rely on the rationale that a coupled dictionary estimated from observed images tends to produce stronger reconstruction errors in change regions rather than in unchanged ones.…”

Learnable Gated Convolutional Neural Network for Semantic Segmentation in Remote-Sensing Images

“…Accordingly, extracting objects of interest in RS images with high quality has become an urgent need. Technically, efforts in semantic segmentation for RS images have surged in recent decades because it is a fundamental approach to analyze RS images, which has many important real-world applications [1][2][3][4], typically including plant-disease detection [5][6][7][8][9][10], land-cover planning [11], vegetation extraction [12], cloud detection [13,14], urban-change detection [15][16][17], vehicle detection [18], building extraction [19,20] and road extraction [21,22]. However, none of these applications reaches satisfactory segmentation quality.…”