Image Correction and In Situ Spectral Calibration for Low-Cost, Smartphone Hyperspectral Imaging

Round 1

Reviewer 1 Report

This paper is used to present the ultra-low-cost system described by Stuart et al. (Journal of Imaging 2021, 7, (8), 136) and point out the existing limitations of the manual scanning technique before the presentation an innovative solution to the problem and demonstrating the potential of a manual hyperspectral image recorder in series.

However,  some clarifications are needed:

• L215, is not the spectral calibration reference target also present in other figures (such as in Fig 7C..)
• Page 11 is blank.
• Figure 10: why exactly those wavelengths

 

You have not indicated the type of the paper. 

Sincerely

Author Response

Thank you for your time and feedback.

1 L215, is not the spectral calibration reference target also present in other figures (such as in Fig 7C..)

The spectral reference target is consistent throughout the figures. In figure 10 (now 11) it appears different because it is a photograph of the target that has been printed. This concern was not shared by the other reviewers and so the authors believe this is sufficiently clear.

2 Page 11 is blank.

Page 11 has been removed.

3 Figure 10: why exactly those wavelengths

Figure 10 (now 11) is to show the instrument's utility as a hyperspectral imager and so wavelengths were chosen to cover the wavelength range of the instrument to highlight how features of the apple are more or less apparent at different wavelengths. Line 355 edited for clarity.

Reviewer 2 Report

Dear authors,

I’ve read your paper focusing on the development of a low-cost smartphone hyperspectral imaging device. The authors focus their research in image calibration and correction and test the instrument with different kind of targets. This paper show that, also using simple RGB dataset, useless information can be extracted and can be efficient to improve different monitoring practices.

I suggest accepting the paper after a MINOR revision.

 

I suggest only few modifications which are listed below.

1. In the introduction authors could add some other different examples related to agricultural and environmental monitoring application (such as civil engineering). This section could be expanded.
2. In materials and methods authors must specify which kind of hyperspectral device was used. Add spectral characteristics such as number of bands an FWHM.
3. Specify the number of acquisitions which have been made for each kind of target.
4. There is a lack on the validation step: no data referring to accuracy o RMSE are provided for both image correction and calibration.

Author Response

Thank you for your time and feedback.

1 In the introduction authors could add some other different examples related to agricultural and environmental monitoring application (such as civil engineering). This section could be expanded.

 

Civil engineering references added to the introduction. Lines 29-33 added to introduction.

2 In materials and methods authors must specify which kind of hyperspectral device was used. Add spectral characteristics such as number of bands an FWHM.

Lines 136-141 added giving reference to the system and specifications.

3 Specify the number of acquisitions which have been made for each kind of target.

Line 315 added stating number of acquisitions.

4 There is a lack on the validation step: no data referring to accuracy o RMSE are provided for both image correction and calibration.

Lines 320-322 stating RMSE values for spectral calibration. The spatial correction is to provide context for the user as stated in lines 286-291 and so not quantitative assessment of spatial correction was made.

Reviewer 3 Report

The submitted work tries to address an important issue in portable hyperspectral imaging: “Vertical/horizontal spatial distortions introduced into images by operator shake”.

I found this work very interesting and a nice follow-up study for the article ‘Low-Cost Hyperspectral Imaging with a Smartphone’ published in 2021 by the same authors in the MDPI Journal of Imaging.

Generally speaking, the manuscript is well-written and well-structured, the topic is interesting, the innovation level is adequate, and the experimental results are enough to convince the potential reader about the performance of the proposed method. Then I have almost no major comments against the publication of such a nice work, although some minor issues are still here to be addressed and solved. I do suggest proofreading the manuscript, since I found some typos here and there. Some of the figures also need to be reproduced because of their relatively low quality. Please also be consistent with the font size and style, as some fonts in the legends of figures (or in some formulas) are super large or in a different style.

Author Response

Thank you for your time and feedback.

Typos have been corrected. Figures 7 (now 8), 9 (now 10), and 10 (now 11) have been corrected for font style and size as per recommendations.

Reviewer 4 Report

The presented work "Image correction and in-situ spectral calibration for low-cost, smartphone hyperspectral imaging" is written quite qualitatively and is devoted to an interesting topic of hyperspectral visualization. I liked the work, and I have almost no comments. As a problem place of all work, I just want to point out the lack of some technical aspects that the work lacks. 1) There is no indication of how it is necessary to scan the space, how this is related to the size of the scale ruler, what frame rate is possible, etc. Focal length of the lens, etc. 2) In Figure 4, distortions in the form of grease are clearly visible. What was the scanning speed at the same time? what is the maximum scanning speed for such an image acquisition method?
3) The biggest drawback of the work is that it is not described exactly how the correction will be carried out if, in addition to the offset during scanning, there are small turns caused by turning the hand or shaking?

It is necessary to supplement the text of the article with answers to these questions.

Author Response

Thank you for your time and feedback.

1 There is no indication of how it is necessary to scan the space, how this is related to the size of the scale ruler, what frame rate is possible, etc. Focal length of the lens, etc. 

Figure 1 (now 2) shows a rendered demonstration of how a scan would be performed relative to the calibration card and the smartphone.

There is no scale ruler, the top section of the calibration card is used for horizontal spatial correction as detailed in lines 190-214.

The framerate of the smartphone’s camera is now stated in line 144.

Typical scan duration is now stated in lines 147-148.

Lines 136-141 give a more detailed reference to the instruments specifications.

2 In Figure 4, distortions in the form of grease are clearly visible

Distortions visible in the images are from operator shake as described in lines 68-76. Lack of spatial resolution is a limitation of the instrument described in lines 295-297.

What was the scanning speed at the same time? what is the maximum scanning speed for such an image acquisition method?

Scan duration is now stated in lines 147-148. Maximum scanning speed is limited by the instrument and is not discussed because it is beyond the scope of this work. Further details of the instruments are detailed in lines 136-141.

3 The biggest drawback of the work is that it is not described exactly how the correction will be carried out if, in addition to the offset during scanning, there are small turns caused by turning the hand or shaking?

Distortions due to slight rotations are present but negligible relative to the vertical and horizontal distortions. Lines 82-84 added. The spatial correction is to provide context for the user as stated in lines 286-291 and so the correction in the horizontal and vertical dimensions are sufficient to provide this.