15:30 〜 15:45
[PPS05-01] Image restoration experiment for infrared images of Venus
キーワード：Venus、Image Enhancement、Akatsuki、Infrared image、BM3D
Thermal infrared images captured by Longwave Infrared Camera (LIR) onboard Akatsuki provides valuable information about the dynamics of Venusian atmosphere. LIR successfully detects thermal emission in a waveband 8-12 μm from the cloud deck of Venus on dayside and nightside with equal quality[1]. However, the noise in the infrared images prevents from further investigation of the Venusian atmosphere. Enhancement of the image quality is thereby needed to reveal more detailed morphology and structures of clouds.
Image restoration has been a popular topic and extensively studied for decades. It is the task of recovering a true unknown image from a degraded observed one. Numerous traditional and deep learning methods have been developed and proven to be powerful in image restoration for fields like remote sensing[2], medical imaging[3], etc. Regarding the LIR images, previous study reduced random noise by taking a moving average of successive images with an interval of a few hours[4,5]. Stable structures are thereby highlighted, while transient ones are suppressed simultaneously.
Therefore, this study seeks new approaches to enhance the remote sensing images while refrain the transient structures from reduction. The Block-Matching and 3D Filtering (BM3D) algorithm is employed to restore the infrared images of Venus, facilitating the identification and tracking of cloud structures. It is a non-local, adaptive, nonparametric filtering methodology based on a sparse representation in transform domain, combining the advantages of spatial filtering and frequency filtering to achieve better denoising effect[6]. The enhanced images reveal more distinguishable cloud structures, which are compared with observations from other wavebands. This study demonstrates the effectiveness of the denoising algorithm BM3D on the LIR images, providing insights for the future image process routine of remote sensing data on Venus. Additionally, this study further analyzes the development of mesoscale cloud structures, contributing to a deeper understanding of Venus's atmospheric dynamics.
[1] Taguchi, M., Fukuhara, T., Imamura, T., Nakamura, M., Iwagami, N., Ueno, M., Suzuki, M., Hashimoto, G. L., and Mitsuyama, K. (2007). Longwave Infrared Camera onboard the Venus Climate Orbiter. Advances in Space Research, 40(6), 861-868.
[2] B. Rasti, Y. Chang, E. Dalsasso, L. Denis, and P. Ghamisi. Image restoration for remote sensing: Overview and tool- box. IEEE Geoscience and Remote Sensing Magazine, 10(2):201–230, 2022.
[3] Jiri Jan. Medical image processing, reconstruction and restoration. Boca Raton, Fl: Taylor Francis, 2006.
[4] Fukuya, K., Imamura, T., Taguchi, M., Fukuhara, T., Kouyama, T., Horinouchi, T., Peralta, J., Futaguchi, M., Yamada, T., Sato, T. M., Yamazaki, A., Murakami, S.-Y., Satoh, T., Takagi, M., and Nakamura, M. (2021). The nightside cloud-top circulation of the atmosphere of Venus. Nature, 595(7868), 511-515.
[5] Fukuya, K., Imamura, T., Taguchi, M., and Kouyama, T. (2022). Horizontal structures of bow-shaped mountain wave trains seen in thermal infrared images of venusian clouds taken by Akatsuki LIR. Icarus, 378, 114936.
[6] Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian (2007). Image denoising by sparse 3-d transform- domain collaborative filtering. IEEE Transactions on Image Processing, 16(8), 2080–2095.
Image restoration has been a popular topic and extensively studied for decades. It is the task of recovering a true unknown image from a degraded observed one. Numerous traditional and deep learning methods have been developed and proven to be powerful in image restoration for fields like remote sensing[2], medical imaging[3], etc. Regarding the LIR images, previous study reduced random noise by taking a moving average of successive images with an interval of a few hours[4,5]. Stable structures are thereby highlighted, while transient ones are suppressed simultaneously.
Therefore, this study seeks new approaches to enhance the remote sensing images while refrain the transient structures from reduction. The Block-Matching and 3D Filtering (BM3D) algorithm is employed to restore the infrared images of Venus, facilitating the identification and tracking of cloud structures. It is a non-local, adaptive, nonparametric filtering methodology based on a sparse representation in transform domain, combining the advantages of spatial filtering and frequency filtering to achieve better denoising effect[6]. The enhanced images reveal more distinguishable cloud structures, which are compared with observations from other wavebands. This study demonstrates the effectiveness of the denoising algorithm BM3D on the LIR images, providing insights for the future image process routine of remote sensing data on Venus. Additionally, this study further analyzes the development of mesoscale cloud structures, contributing to a deeper understanding of Venus's atmospheric dynamics.
[1] Taguchi, M., Fukuhara, T., Imamura, T., Nakamura, M., Iwagami, N., Ueno, M., Suzuki, M., Hashimoto, G. L., and Mitsuyama, K. (2007). Longwave Infrared Camera onboard the Venus Climate Orbiter. Advances in Space Research, 40(6), 861-868.
[2] B. Rasti, Y. Chang, E. Dalsasso, L. Denis, and P. Ghamisi. Image restoration for remote sensing: Overview and tool- box. IEEE Geoscience and Remote Sensing Magazine, 10(2):201–230, 2022.
[3] Jiri Jan. Medical image processing, reconstruction and restoration. Boca Raton, Fl: Taylor Francis, 2006.
[4] Fukuya, K., Imamura, T., Taguchi, M., Fukuhara, T., Kouyama, T., Horinouchi, T., Peralta, J., Futaguchi, M., Yamada, T., Sato, T. M., Yamazaki, A., Murakami, S.-Y., Satoh, T., Takagi, M., and Nakamura, M. (2021). The nightside cloud-top circulation of the atmosphere of Venus. Nature, 595(7868), 511-515.
[5] Fukuya, K., Imamura, T., Taguchi, M., and Kouyama, T. (2022). Horizontal structures of bow-shaped mountain wave trains seen in thermal infrared images of venusian clouds taken by Akatsuki LIR. Icarus, 378, 114936.
[6] Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian (2007). Image denoising by sparse 3-d transform- domain collaborative filtering. IEEE Transactions on Image Processing, 16(8), 2080–2095.