Column-Spatial Correction Network for Remote Sensing Image Destriping

Abstract

The stripe noise in the multispectral remote sensing images, possibly resulting from the instrument instability, slit contamination, and light interference, significantly degrades the imaging quality and impairs high-level visual tasks. The local consistency of homogeneous region in striped images is damaged because of the different gains and offsets of adjacent sensors regarding the same ground object, which leads to the structural characteristics of stripe noise. This can be characterized by the increased differences between columns in the remote sensing image. Therefore, the destriping can be viewed as a process of improving the local consistency of homogeneous region and the global uniformity of whole image. In recent years, convolutional neural network (CNN)-based models have been introduced to destriping tasks, and have achieved advanced results, relying on their powerful representation ability. Therefore, to effectively leverage both CNNs and the structural characteristics of stripe noise, we propose a multi-scaled column-spatial correction network (CSCNet) for remote sensing image destriping, in which the local structural characteristic of stripe noise and the global contextual information of the image are both explored at multiple feature scales. More specifically, the column-based correction module (CCM) and spatial-based correction module (SCM) were designed to improve the local consistency and global uniformity from the perspectives of column correction and full image correction, respectively. Moreover, a feature fusion module based on the channel attention mechanism was created to obtain discriminative features derived from different modules and scales. We compared the proposed model against both traditional and deep learning methods on simulated and real remote sensing images. The promising results indicate that CSCNet effectively removes image stripes and outperforms state-of-the-art methods in terms of qualitative and quantitative assessments.

Wavelet Channel Attention Module With A Fusion Network For Single Image Deraining

“…Guan et al [23] designed a cascade residual attention CNN model that estimates the gain and bias of the noise response using two concatenated sub-networks. Taking into account the signal readout by column, Li et al [30] designed a column-based stripe correction attention module to ensure local consistency in uniform regions and global uniformity in the image. Different from them, our destriping method leverages structural properties using spatial attention, column attention, and selfcalibration.…”

“…Different from them, our destriping method leverages structural properties using spatial attention, column attention, and selfcalibration. Specifically, our technique differs from [30] along the following lines. We calculate the average and maximum response of the columns by fusing the different response features, and calibrate the local noise non-uniformity in two steps.…”

“…Therefore, destriping task can be translated into improving local consistency within homogeneous regions and attaining global image uniformity. In previous work [30], a column-spatial correction module was proposed to mitigate the column-to-column differences. However, the local nonuniformity between columns is not adequately characterized due to premature expansion of the feature map, resulting in rapid decay of the inter-column correlation and causing localized streak residue.…”

“…However, using these samplers often produce varying degrees of semantic and structural information loss, leading to ineffective feature representations. Likewise, due to their lower inductive bias, attention-based destriping methods [23], [30] fail to fully exploit the characteristic column patterns of the stripe noise.…”

HSI-IPGAN: Hyperspectral Image Inpainting via Generative Adversarial Network

“…However, HSIs often lose certain image information with different shapes caused by the instability of the hyperspectral imaging system and atmospheric interference. For example, the detector damage leads to the striped missing areas and cloud occlusion results in irregular-shaped missing areas in HSIs, which seriously limit the effectiveness of downstream tasks, such as classification [1,[5][6][7], destriping [8,9], and detection [3,10] etc.. Therefore, HSI inpainting plays a crucial role in remote sensing image processing to further improve the visual quality and promote the subsequent applications.…”