Super-Resolution Reconstruction of Remote Sensing Images Using Multiple-Point Statistics and Isometric Mapping

Abstract

When using coarse-resolution remote sensing images, super-resolution reconstruction is widely desired, and can be realized by reproducing the intrinsic features from a set of coarse-resolution fraction data to fine-resolution remote sensing images that are consistent with the coarse fraction information. Prior models of spatial structures that encode the expected features at the fine (target) resolution are helpful to constrain the spatial patterns of remote sensing images to be generated at that resolution. These prior models can be used properly by multiple-point statistics (MPS), capable of extracting the intrinsic features of patterns from prior models such as training images, and copying them to the simulated regions using hard and soft conditional data, or even without any conditional data. However, because traditional MPS methods based on linear dimensionality reduction are not suitable to deal with nonlinear data, and isometric mapping (ISOMAP) can reduce the dimensionality of nonlinear data effectively, this paper presents a sequential simulation framework for generating super-resolution remote sensing images using ISOMAP and MPS. Using four different examples, it is demonstrated that the structural characteristics of super-resolution reconstruction of remote sensing images using this method, are similar to those of training images.

Research on super-resolution reconstruction of remote sensing images: a comprehensive review

“…Simultaneously, by considering that abnormal values are often included in the embedding and that they reduce the structure accuracy, they used robust sparse embedding to eliminate outliers and normalized the weights to obtain more accurate neighborhoods and coding coefficients when synthesizing HR images. Zhang et al 43 considered that multiple-point simulation based on linear dimensionality reduction compresses a high-dimensional space and shortened the simulation time, but it also reduces the simulation quality and limits the use for applications to non-linear data. Therefore, they introduced isometric mapping to achieve non-linear dimensionality reduction and combined multiple-point simulation with clustering for classification after dimensionality reduction.…”

LIDAR Point Cloud Registration for Sensing and Reconstruction of Unstructured Terrain

“…where d ij is the spatial distance from one point to its neighbor point, k j is the curvature of the neighbor point, f j is the neighborhood weighted angle-invariant feature, m is the number of neighbor points. It should be noticed that x, y, z, f, k, u, and v constitute our feature descriptor, in which x, y, z are the coordinates of point, f is the neighborhood weighted angle-invariant feature, k is the curvature, u and v are from formula (6). We use the data obtained by calculating the normal vectors of points to build a feature that fully expresses the neighborhood information, thus the speed and accuracy of feature calculation and correspondence building are improved.…”

“…Therefore, various sensors such as visual image sensors [1,2] and LIDAR [3,4] have been deployed on autonomous vehicles for sensing unstructured terrain to achieve safe operation. Although the visual image sensors only produce 2D images, sufficient terrain information can be obtained by using some approaches, such as calculating a height map from an image to obtain 3D information [5] or using the super-resolution reconstruction method to obtain a spatial resolution-enhanced image [6]. Compared with visual image sensors, LIDAR has the advantage of being free from the effect of light and weather and can obtain the three-dimensional terrain information conveniently.…”