Sparse Constrained Low Tensor Rank Representation Framework for Hyperspectral Unmixing
Round 1
Reviewer 1 Report
The authors propose a novel method based on non-negative tensor factorization for hyperspectral unmixing, which combines the multi-mode low-rank and sparse structure of the abundance tensor. The proposed approach is evaluated on three real data with several state-of-the-art comparison methods and the results reveal the superiority of the method. In addition, this paper has a good paper organization, writing and rich figures to illustrate the experimental results.
Following are some suggestions for the authors:
- 1. The font on the left side of Fig. 7 is too small.
- 2. Fig. 2, 4, and 6 should give the name of each endmember spectrum.
- It is recommended to appropriately increase the references published on remote sensing. For example,
[1] Kizel F , Jón Atli Benediktsson. Spatially Enhanced Spectral Unmixing Through Data Fusion of Spectral and Visible Images from Different Sensors. Remote Sensing, 2020, 12(8):1255.
[2] Zheng, Feiyang W U , Shim, et al. Sparse Unmixing for Hyperspectral Image with Nonlocal Low-Rank Prior. Remote Sensing, 2019, 11(24):2897.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 2 Report
Please see attached file.Comments for author File: Comments.pdf
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper presents a spectral unmixing method with sparse low-rank tensor representations. Overall, the structure of this paper is well organized. However, there are still some crucial issues that need to be carefully considered before a possible publication. More specifically,
- Sparse and low-rank modeling is nothing new in spectral unmixing. It would be better to clarify the newly-added values in this paper.
- A deep literature review related to the hyperspectral image processing should be given, particularly regarding spectral unmixing, in order to show state-of-the-arts of the proposed method. Therefore, the reviewer strongly suggests discussing and analyzing latest advanced and latest works, e.g., “An Augmented Linear Mixing Model to Address Spectral Variability for Hyperspectral Unmixing, IEEE Transactions on Image Processing, 2019, 28(4): 1923-1938”, “Nonconvex-Sparsity and Nonlocal-Smoothness-Based Blind Hyperspectral Unmixing, IEEE Transactions on Image Processing, 2019, 28(6): 2991-3006”, “Endmember extraction of hyperspectral remote sensing images based on the ant colony optimization (ACO) algorithm. IEEE transactions on geoscience and remote sensing, 2011, 49(7), 2635-2646.”, and “Graph Convolutional Networks for Hyperspectral Image Classification, IEEE Transactions on Geoscience and Remote Sensing, 2020, DOI: 10.1109/TGRS.2020.3015157.”
- In the experiments, why choose the indinepine as the dataset for spectral unmixing? How to evaluate it, since the raw data have 16 classes.
- How about the computational cost, since low-rank tensor-based model is usually expensive in computational cost.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
No further comments.
Reviewer 3 Report
The authors have well addressed the reviewer's concerns. No more comments!