Askaniy
Titan Enhanced Color Map (11K)
Saturn's moon Titan is one of the most difficult objects to figure out true colors. To begin with, its surface is simply impossible to observe in the visible range because of the haze. Sufficiently transparent spectral bands begin only in the infrared, which means that only for the infrared it is possible to create a global map. Fortunately, we have Huygens landing probe, whose data was processed by Erich Karkoschka and Stefan E. Schröder [1]. The figure 6 was run through my program TrueColorTools (requiring some modifications along the way). The multiband images were interpolated and extrapolated (to blue range) there to a spectral cube and then convolved with the sensitivity of the human eye.
Having surface panoramas with a known color processing, all that is left is to extend that color to some infrared map. The best available option was a map from B. Seignovert et al. [2], which was first rid of artifacts and manually cleaned of pixilization. The Huygens landing site was selected from this map and matched to a previously processed projection of it in visible colors.
Now we need to find an unknown transformation over a piece of the infrared map so that it produces something as close as possible to the corresponding piece of the visible map, and then apply the found transformation to the entire infrared map. Assuming linearity of the transformation, it can be written in the form IR · X + C = VIS, where IR is a known 3-vector of some pixel color, X is an unknown 3x3 color transformation matrix, C is an unknown 3-vector and VIS is the resulted visible color 3-vector. The linearity assumption can be justified if the surface spectra in the visible and infrared are highly correlated with each other (which is true), but the infrared map used [2] contains nonlinear color transformations to highlight geologic features. Therefore, it was renormalized as follows before processing: R′ = G · B = 2.03/1.08 μm, G′ = R · B = 1.59/1.08 μm, B′ = B = 1.27/1.08 μm.
It turns out that we do not know the 3x3+3=12 parameters of X and C responsible for color conversion. They were found in 12-dimensional space as a result of optimization over all the landing site pieces pixels by the MNC method in Python. There was an attempt to add another matrix with quadratic form, but it failed: the result was too optimized and unrealistic.
Unfortunately, Huygens was only able to capture a small region of the surface, only one of two variations in the coloration of the dunes. Therefore, the unknown color of the second type of dunes had to be handpicked based on the assumption that there were no sharp brightness gradients on the surface, and run the optimization with this patch of handpicked influence. The color of the lakes is also handpicked, no suitable theoretical data has been found. Without these assumptions, the texture map cannot be completed. That's about as close to maximum color realism as we can achieve right now.
Lake delineations were obtained separately using radar data [3], which had noise and missing data. These regions were manually reconstructed from infrared data from Cassini, and the infrared map [2] itself was pre-warped to match the radar data (reprojected from the polar stereographic projection with another Python script).
Thanks to Pedro J. and Chara for their help and support!
History
August 2025: the Titan color map series was updated in accordance with the color processing updates in TrueColorTools (Tikhonov regularization for spectral reconstruction, color spaces management).
Contrast is slightly increased due to the darker presumed visible color of the second type of dunes.
January 2026: Did a little research on the color of liquid methane and ethane under Titan surface conditions. Turns out they're likely transparent, see this paper. The spherical albedo color was calculated from the refractive index "n" from [4] by integrating the Fresnel equations. The script makes the average color of the lakes calculated taking into account map distortions.
Info
Simple cylindrical projection, center longitude 0°.
Gamma corrected, albedo corrected.
Sources
[1] Karkoschka et al. (2016). Eight-color maps of Titan’s surface from spectroscopy with Huygens’ DISR
[2] B. Seignovert et al. (2019). Titan's global map combining VIMS and ISS mosaics (1.1). CaltechDATA
[4] Martonchik & Orton (1994). Optical constants of liquid and solid methane
Related
Titan Enhanced Color Map (11K)
Saturn's moon Titan is one of the most difficult objects to figure out true colors. To begin with, its surface is simply impossible to observe in the visible range because of the haze. Sufficiently transparent spectral bands begin only in the infrared, which means that only for the infrared it is possible to create a global map. Fortunately, we have Huygens landing probe, whose data was processed by Erich Karkoschka and Stefan E. Schröder [1]. The figure 6 was run through my program TrueColorTools (requiring some modifications along the way). The multiband images were interpolated and extrapolated (to blue range) there to a spectral cube and then convolved with the sensitivity of the human eye.
Having surface panoramas with a known color processing, all that is left is to extend that color to some infrared map. The best available option was a map from B. Seignovert et al. [2], which was first rid of artifacts and manually cleaned of pixilization. The Huygens landing site was selected from this map and matched to a previously processed projection of it in visible colors.
Now we need to find an unknown transformation over a piece of the infrared map so that it produces something as close as possible to the corresponding piece of the visible map, and then apply the found transformation to the entire infrared map. Assuming linearity of the transformation, it can be written in the form IR · X + C = VIS, where IR is a known 3-vector of some pixel color, X is an unknown 3x3 color transformation matrix, C is an unknown 3-vector and VIS is the resulted visible color 3-vector. The linearity assumption can be justified if the surface spectra in the visible and infrared are highly correlated with each other (which is true), but the infrared map used [2] contains nonlinear color transformations to highlight geologic features. Therefore, it was renormalized as follows before processing: R′ = G · B = 2.03/1.08 μm, G′ = R · B = 1.59/1.08 μm, B′ = B = 1.27/1.08 μm.
It turns out that we do not know the 3x3+3=12 parameters of X and C responsible for color conversion. They were found in 12-dimensional space as a result of optimization over all the landing site pieces pixels by the MNC method in Python. There was an attempt to add another matrix with quadratic form, but it failed: the result was too optimized and unrealistic.
Unfortunately, Huygens was only able to capture a small region of the surface, only one of two variations in the coloration of the dunes. Therefore, the unknown color of the second type of dunes had to be handpicked based on the assumption that there were no sharp brightness gradients on the surface, and run the optimization with this patch of handpicked influence. The color of the lakes is also handpicked, no suitable theoretical data has been found. Without these assumptions, the texture map cannot be completed. That's about as close to maximum color realism as we can achieve right now.
Lake delineations were obtained separately using radar data [3], which had noise and missing data. These regions were manually reconstructed from infrared data from Cassini, and the infrared map [2] itself was pre-warped to match the radar data (reprojected from the polar stereographic projection with another Python script).
Thanks to Pedro J. and Chara for their help and support!
History
August 2025: the Titan color map series was updated in accordance with the color processing updates in TrueColorTools (Tikhonov regularization for spectral reconstruction, color spaces management).
Contrast is slightly increased due to the darker presumed visible color of the second type of dunes.
January 2026: Did a little research on the color of liquid methane and ethane under Titan surface conditions. Turns out they're likely transparent, see this paper. The spherical albedo color was calculated from the refractive index "n" from [4] by integrating the Fresnel equations. The script makes the average color of the lakes calculated taking into account map distortions.
Info
Simple cylindrical projection, center longitude 0°.
Gamma corrected, albedo corrected.
Sources
[1] Karkoschka et al. (2016). Eight-color maps of Titan’s surface from spectroscopy with Huygens’ DISR
[2] B. Seignovert et al. (2019). Titan's global map combining VIMS and ISS mosaics (1.1). CaltechDATA
[4] Martonchik & Orton (1994). Optical constants of liquid and solid methane
Related