Systematic Approach for Alignment of Light Field Mirage

Round 1

Reviewer 1 Report

In this paper, the authors develop an effective alignment technique for the light field Mirage for a complete 360-degree display system. They propose a systematic approach for the alignment for the light field Mirage, which causes less image quality degradation by taking advantage of the small number of display devices required for the light field Mirage. The calibration technique for light field displays, the image stitching technique, and the brightness matching technique are consecutively performed and the generation of 360-degree 3D images is verified. It is based on previously proposed techniques to eliminate repeated three-dimensional (3D) images produced by the light field Mirage, which consists of circularly-aligned multiple-slanted light field displays. It has improvement compared with previous method. Overall, I recommend receiving as long as modifying the following problems:

1、 Compared with previous method, why it can be extended to 360-degree? What’s the difference between the two methods? I hope the authors can explain it in details.

2、 There are some improper subsections, such as in line 65, line 120, a sentence may not be divided as a single section.

3、 In Figs.9 and 10, what are the differences between a and b? If there are no major differences, they can be combined together. There's no need to take up so much space.

4、 It is recommended that all images are vector images, so as not to lose resolution.

Author Response

Please check the attached file for details.

Response to Reviewer 1

[Overall Comment] In this paper, the authors develop an effective alignment technique for the light field Mirage for a complete 360-degree display system. They propose a systematic approach for the alignment for the light field Mirage, which causes less image quality degradation by taking advantage of the small number of display devices required for the light field Mirage. The calibration technique for light field displays, the image stitching technique, and the brightness matching technique are consecutively performed and the generation of 360-degree 3D images is verified. It is based on previously proposed techniques to eliminate repeated three-dimensional (3D) images produced by the light field Mirage, which consists of circularly-aligned multiple-slanted light field displays. It has improvement compared with previous method. Overall, I recommend receiving as long as modifying the following problems:

[Answer] We thank the reviewer for reading our manuscript carefully and providing us with valuable comments. We have revised our manuscript in accordance with the comments as described in the following.

[Comment 1] Compared with previous method, why it can be extended to 360-degree? What’s the difference between the two methods? I hope the authors can explain it in details. 

[Answer] As the reviewer pointed out, the explanation in the original manuscript was confusing. We have revised our manuscript to address this confusion.

First paragraph in Sec. 1:

... However, we only constructed a part of the system to verify the elimination techniques. Thus, the previous system could not achieve 360-degree visibility. In this study, we develop a technology that connects multiple light field displays (LFDs) to complete the 360-degree visibility, which includes calibration for LFDs, stitching of images in overlapped LFDs, and matching of LFD’s brightness.

[Comment 2] There are some improper subsections, such as in line 65, line 120, a sentence may not be divided as a single section.

[Answer] We have also revised our manuscript in reference to the reviewer’s comment.

First paragraph in Sec. 2:

First, we briefly explain the light field Mirage that was proposed in our previous study [18].We previously proposed a concept of digital implementation system in which LFDs, which substitute the existing Mirage parabolic mirrors, generate a 360-degree 3D image in the top hole as shown in Fig. 1…

First paragraph in Sec. 3:

Prior to the development of the alignment techniques, we completed the design of the full light field Mirage. We provided the same LFDs as those used in our previous study [18]…

[Comment 3] In Figs.9 and 10, what are the differences between a and b? If there are no major differences, they can be combined together. There's no need to take up so much space.

[Answer] The 3D result images have also been combined following the reviewer’s comment.

First paragraph in Sec. 7:

…Figure 9(a) shows photographs of the 3D images captured from 16 directions. The eight directions (0.0°, 45.0°, 90.0°, 135.0°, 180.0°, 225.0°, 270.0°, and 315.0°) were the directions normal to the screens of the eight LFDs and the other eight directions (22.5°, 67.5°, 112.5°, 157.5°, 202.5°, 247.5°, 292.5°, and 337.5°) were the intermediate directions of the former eight directions. The magnified images are shown in Fig. 9(b). Because no double images were observed from any direction, the calibration technique worked well…

Figure 9. 360-degree 3D image generated by non-tracking technique: (a) observed from 16 directions and (b) magnified 3D images.

Second paragraph in Sec. 7:

…The displayed objects were chess pieces: a knight and pawn. Figure 10(a) shows photo-graphs of the 3D images captured from the 16 directions used for the experiment of the non-tracking technique. The magnified images are also shown in Fig. 10(b)...

Figure 10. 360-degree 3D image generated by tracking technique: (a) observed from 16 directions. and (b) magnified 3D images.

[Comment 3] It is recommended that all images are vector images, so as not to lose resolution.

[Answer] Thank you for your advice. Images were revised as vector graphic data such as the Enhanced Metafile format (".EMF").

Author Response File: Author Response.docx

Reviewer 2 Report

In this paper, the authors propose a systematic approach for creating high-quality 3D images based on multi-projection and flat-panel systems. Both tracking and non-tracking architectures are developed to eliminate repeated images. The authors also develop systematic alignment techniques to reduce image mismatch, stitching errors and brightness nonuniformity.

Overall, the proposed techniques are practical in 360-deg 3D displays. Although similar systems have been demonstrated in previous studies (Ref [18]), this work extends this approach for a complete 360-deg display. Thus, I recommend acceptance of the manuscript. Regarding the contents, I have some comments below:

1.     Full acronym for “LFD” (light field display?) should be specified in the paper.

2.    On page 4, the authors state that the reason to use metallic half mirror is because “large” mirror is required. This is confusing since dielectric mirrors can also be made at a large scale. The authors should elaborate on that.

3.     Despite the demonstration of 360-deg images, could the authors also comment on chromatic aberration of the system? Besides, the wavelengths of light used in the experiments (e.g. Fig. 4,5,9,10) are recommended to be specified.

4.     In Fig.9 and 10, it seems that the images at directions normal to the screens (e.g. 0°, 90°) exhibit higher resolution than the intermediate angles (e.g. 157). More discussion on this is recommended.

Author Response

Please check the attached file for details.

Response to Reviewer 2

[Overall Comment] In this paper, the authors propose a systematic approach for creating high-quality 3D images based on multi-projection and flat-panel systems. Both tracking and non-tracking architectures are developed to eliminate repeated images. The authors also develop systematic alignment techniques to reduce image mismatch, stitching errors and brightness nonuniformity.

Overall, the proposed techniques are practical in 360-deg 3D displays. Although similar systems have been demonstrated in previous studies (Ref [18]), this work extends this approach for a complete 360-deg display. Thus, I recommend acceptance of the manuscript. Regarding the contents, I have some comments below:

[Answer] Thank you very much for carefully reading our manuscript. We have revised our manuscript following the reviewer’s comments.

[Comment 1]Full acronym for “LFD” (light field display?) should be specified in the paper.

[Answer] As the reviewer pointed out, the full acronym for light field display (LFD) had not specified. Thank you very much for the kind suggestion. The manuscript has been revised following the reviewer’s comment.

First paragraph in Sec. 1 (underlined):

Observing 360-degree three-dimensional (3D) display systems can be used for digital signage, medical demonstrations, and Internet shopping. We previously proposed techniques to eliminate repeated 3D images produced by the light field Mirage, which produces 360-degree 3D images to simulate the conventional Mirage [18]. However, we only constructed the lower half of the system to verify the elimination techniques. Thus, the previous method could not achieve 360-degree visibility. In this study, we develop a technology that connects multiple light field displays (LFDs) to complete the 360-degree visibility, which include calibration for LFDs, stitching of images in overlapped LFDs, and matching of LFD’s brightness.

[Comment 2] On page 4, the authors state that the reason to use metallic half mirror is because “large” mirror is required. This is confusing since dielectric mirrors can also be made at a large scale. The authors should elaborate on that.

[Answer] Thank you very much for your suggestion. The manuscript has been revised following the reviewer’s comment.

Third paragraph in Sec. 3:

...Although the light absorption of the metallic half mirrors is higher than that of dielectric half mirrors, we used a metallic half mirror because a large dielectric half mirror requires high costs and has strict restrictions on wavelengths and incident angles...

[Comment 3] Despite the demonstration of 360-deg images, could the authors also comment on chromatic aberration of the system? Besides, the wavelengths of light used in the experiments (e.g. Fig. 4,5,9,10) are recommended to be specified.

[Answer] In section 4, we conducted a calibration experiment by using different colors for each display to easily identify the shifted display. Therefore, if the images are not superimposed on one another, the color will be shifted. The result shown in Fig. 5 (b) was colored because a part of the displays was hindered by the top hole. The wavelength of light used in the experiments was unknown because the specifications of the LCD were not written. Nevertheless, it was normal a LCD. 3D images can be susceptible to chromatic aberration by using a plastic lens. However, we did not conduct color correction in the experiment (e.g. Figs. 9 and 10) .

The explanation in the original manuscript was confusing. We have revised our manuscript to address this confusion.

Third paragraph in Sec. 4:

In the second step, the positions of the eight LFDs were adjusted. The LFDs were attached to the upper and lower aluminum plates using off-the-shelf 45-degree angle brackets. A sheet of tracing paper was placed at the top hole of the light field Mirage and each LFD displayed a grid pattern at the top hole position (the CDP position). The positions of the LFDs were moved along the aluminum plates so that all gird patterns were superimposed on one another. The grid pattern used a different color for each LFD to easily identify the shifted LFD. The resultant superimposed image is shown in Fig. 4(a). As shown in this figure, it was impossible to superimpose all grid patterns at the same position, because of the residual non-uniformity among the eight LFDs after the first step. Then, the tracing paper was moved upward over a distance of 14 mm and all the LFDs displayed the grid patterns at the same position at this height. Figure 4(b) shows the superimposition of the grid patterns. The mismatch of the grid pattern positions became larger compared with that shown in Fig. 4(a). Therefore, adjustment for the ray directions among the LFDs was required.

Fourth paragraph in Sec. 4:

In the last step, the electronic calibration was conducted for the adjustment of the ray directions. The positions of the elementary images, as well as, the pitches of the elementary images displayed on the LCDs were manually adjusted so that all grid patterns were superimposed on one another. The adjustments were conducted using the tracing paper positioned at the top hole. The maximum shift of the elementary images among the eight displays was three pixels. From the resultant image shown in Fig. 5(a), the superposition of the grid patterns improved. Then, the tracing paper was moved as explained in the second step and the obtained image is shown in Fig. 5(b). The result shown in Fig. 5 (b) was colored because a part of the displays was hindered by the top hole, but the coincidence of the grid patterns was as good as that shown in Fig. 5(a). Therefore, the ray directions of all the LFDs were well calibrated.

[Comment 4] In Fig.9 and 10, it seems that the images at directions normal to the screens (e.g. 0°, 90°) exhibit higher resolution than the intermediate angles (e.g. 157). More discussion on this is recommended.

[Answer] As the reviewer pointed out, images at directions normal to the screens tend to have higher resolution than the intermediate angles. An explanation has been added in the revised manuscript following the reviewer’s comment.

Added third paragraph in Sec. 7:

As respectively shown in Figs. 9 and 10, in both the non-tracking and tracking methods, 3D images at directions normal to the screens tend to have higher resolution than those at the intermediate angles. The 3D images at the intermediate angles overlapped two LFDs that had CDPs in different planes. The result of viewing angle differences in 3D image resolutions suggest that this is caused because the resolution-prioritized 3D image quality depends on the distance from the CDP.   

Author Response File: Author Response.docx