Improving the Recognizability of Objects on Celestial Surfaces Using Overlap between Imaging Frames from JANUS onboard JUICE

Introduction
 This study aims to improve the recognizability of objects (e.g., craters) in imaging data of the surfaces of celestial bodies acquired by JANUS (Jovis, Amorum ac Natorum Undique Scrutator), a two-dimensional spectral imaging camera on the JUpiter ICy moons Explorer (JUICE)^1,2), a joint mission of the European Space Agency (ESA) and the Japan Aerospace Exploration Agency (JAXA).

 JUICE was launched on April 14, 2023, and is scheduled to reach the Jupiter system in 2031. After multiple flybys of Jupiter, Ganymede, Europa, and Callisto until November 2034, it will perform an orbital observation of Ganymede. The mission is planned to conclude with a collision into Ganymede in 2035. The main scientific targets of JANUS are the geological and topographical features of these icy satellites, as well as phenomena such as lightning occurring in Jupiter’s upper atmosphere. Thus, the imaging data from JANUS covers a wide range of topics. And of course, its higher resolution data offers more advantages in planetary science studies. For example, in crater chronology, higher resolution allows for the identification of more craters, which ultimately enables more accurate model age determinations. Detailed crater morphology also contributes to understanding the impact process, timing, and the physical constraints of the impacted celestial body’s surface. However, due to constraints such as the spacecraft's altitude, the total amount of internal data storage, and the transmission data rate to the Earth, the resolution of the data acquired by JANUS will be strictly limited. This study, therefore, explores the possibility of improving the object recognizability by increasing the resolution in a pseudo manner for JANUS imaging data using frame overlap. The data from the August 2024 Moon and Earth flybys, which included overlapping frames obtained by JANUS, will be used for this experiment. The lunar imaging data, obtained by Japan’s SELENE ^{e.g. 3)} and the U.S. LRO ^{e.g. 4)} missions at high resolutions, will be compared to evaluate the potential for enhancing the JANUS data.

Methodology

 JANUS is a two-dimensional imaging camera with 2000 x 1504 pixels (along track x across track) and a field of view of 1.72° x 1.29°^1,2). The instantaneous field of view (IFOV) of each pixel is 15 μrad, which corresponds to 7.5 m/pixel when observed from an altitude of 500 km (e.g. during the Ganymede flyby). Depending on the operation, JANUS images can be captured in succession to create frame overlap. For this study, we used the overlapping frame data acquired during the JUICE Moon flyby in August 2024^1,2).

 The image enhancement process follows the steps outlined below (see Fig. 1):

1) Sub-Pixel Interpolation: The first step is sub-pixel interpolation to enhance image resolution in a pseudo manner.

2) Image Registration: Image registration is performed using template matching at a sub-pixel level, where the difference in brightness values is minimized.

3) Stacking and Normalization: The aligned overlapping images are stacked, and the radiance values are averaged to normalize the result.

The resulting "processed image" will be compared with high-resolution images from SELENE and LRO to assess the degree of enhancement achieved.

Sub-pixel interpolation will be performed using the Awesome Sindones (AS) processing. This method aims to reduce or remove mosaic artifacts that occur during image enlargement, and is expected to improve the recognizability of features such as shape and texture that are difficult to identify in the original image.

 The key features of the AS processing are as follows:

• It is a quick and easy processing.
• It is reversible, meaning the original image can be reconstructed.
• It can significantly reduce mosaic artifacts with repeated applications.
• It enhances edges such as object contours.

 Among these features, the reversibility of AS processing is particularly important because it allows other researchers to convert the AS processed data back to the original images and apply different preprocessing methods separately. This is not possible with traditional interpolation methods such as bilinear or cubic convolution.

Results
 We present the results of our attempt to process panchromatic data (footprints #50-#60; see Fig. 2) obtained during the JANUS lunar flyby and improve the recognizability of lunar craters. These data were taken on the lunar surface at approximately 9°S and 66°E. Sample images (footprints #61-#63, see Fig. 3) are shown in the figure (filter: FPAN, pixel scale: 18.7-18.9 m/pixel). Comparing the original data with the processed images, we can see that the edges of some craters are enhanced in the processed images, making the details more clearly visible (Fig. 4). Furthermore, comparing the results of the AS method with other interpolation methods (bilinear and cubic convolution methods), we can see that the AS processing enhances edges more effectively, making craters easier to identify (Fig. 5).

Conclusion
 Higher resolution image data of celestial surfaces obtained by spacecraft is desirable to obtain more scientific information. However, it is limited by various constraints. We have demonstrated that applying image processing techniques such as subpixel matching to panchromatic overlapping frame data obtained by JANUS during the JUICE Moon flyby in August 2024 can improve the resolution in a pseudo manner, and for example, significantly improve the recognizability of craters. In particular, Awesome Sindones (AS) processing was effective as a pre-processing step for subpixelization. Unlike conventional interpolation methods (e.g. bilinear and cubic convolution), AS processing is reversible, which is an important advantage. Future research will also explore the possibility of improving the recognizability of observed objects with other types of data, such as JANUS non-panchromatic filtered and compressed image data.

References

1) Palumbo, P. et al. Space Sci. Rev. 221(32), 2025. 2), Lucchettii, A. et al. EPSC, 2025, 3) Haruyama, J. et al. Earth Planets Space 60 (4), 243-256, 2008, 4) Robinson, M. S. et al. Space Sci Rev. 15, 81–124, 2010.