[Paper Review] Hyperspectral Unmixing: Ground Truth Labeling, Datasets, Benchmark Performances and Survey

Feiyun Zhu|arXiv (Cornell University)|Aug 17, 2017

Remote-Sensing Image Classification99 references83 citations

TL;DR

The paper introduces a general ground-truth labeling method for hyperspectral unmixing (HU), summarizes 15 real HU datasets with 18 ground-truth variants, proposes transforming hyperspectral classification data for HU, and provides benchmarks and code for reproducible evaluation of HU methods.

ABSTRACT

Hyperspectral unmixing (HU) is a very useful and increasingly popular preprocessing step for a wide range of hyperspectral applications. However, the HU research has been constrained a lot by three factors: (a) the number of hyperspectral images (especially the ones with ground truths) are very limited; (b) the ground truths of most hyperspectral images are not shared on the web, which may cause lots of unnecessary troubles for researchers to evaluate their algorithms; (c) the codes of most state-of-the-art methods are not shared, which may also delay the testing of new methods. Accordingly, this paper deals with the above issues from the following three perspectives: (1) as a profound contribution, we provide a general labeling method for the HU. With it, we labeled up to 15 hyperspectral images, providing 18 versions of ground truths. To the best of our knowledge, this is the first paper to summarize and share up to 15 hyperspectral images and their 18 versions of ground truths for the HU. Observing that the hyperspectral classification (HyC) has much more standard datasets (whose ground truths are generally publicly shared) than the HU, we propose an interesting method to transform the HyC datasets for the HU research. (2) To further facilitate the evaluation of HU methods under different conditions, we reviewed and implemented the algorithm to generate a complex synthetic hyperspectral image. By tuning the hyper-parameters in the code, we may verify the HU methods from four perspectives. The code would also be shared on the web. (3) To provide a standard comparison, we reviewed up to 10 state-of-the-art HU algorithms, then selected the 5 most benchmark HU algorithms, and compared them on the 15 real hyperspectral datasets. The experiment results are surely reproducible; the implemented codes would be shared on the web.

Motivation & Objective

• Propose a general labeling method for HU to create endmember and abundance ground truths.
• Summarize and share 15 real HU images with 18 ground-truth variants to standardize evaluation.
• Provide a synthetic HU image generation approach and share its code for reproducibility.
• Review and benchmark state-of-the-art HU algorithms on the 15 real HU datasets.

Proposed method

• Label endmembers and abundances for HU using Section IV-A and IV-B methods.
• Provide evaluation framework for labeling results (Section IV-C).
• Transform HyC benchmark datasets into HU-friendly formats (Section IV-D).
• Generate and share a complex synthetic HU image (Section VI).
• Review up to 10 HU methods and implement 5 as benchmark algorithms (Section III).
• Compare the 5 benchmark HU methods on the 15 real HU datasets (Section VII).

Experimental results

Research questions

• RQ1 How can we systematically label endmembers and abundances for HU across multiple datasets?
• RQ2 Can HyC benchmark datasets be transformed to HU-consistent datasets for standardized evaluation?
• RQ3 What is the impact of a complex synthetic HU image for method benchmarking?
• RQ4 Which HU methods provide the best benchmark performance on real HU datasets?
• RQ5 Are the provided codes and ground-truth datasets sufficient for reproducible HU evaluation?

Key findings

• A general labeling method labels endmembers and abundances and provides evaluation for labeling outcomes (Section IV).
• 18 ground-truth variants are created from 15 real HU images, serving as a shared standard dataset for HU evaluation.
• A complex synthetic HU image is generated to enable multi-perspective HU testing and its code is shared.
• The paper reviews 10 HU methods and provides codes for 5 state-of-the-art methods.
• Five benchmark HU methods are implemented and evaluated on the 15 real HU datasets (reproducible results).