Balancing the terminator gradient

Near the terminator, the spherical lunar surface gradually passes into darkness. The image is unbalanced and numerous shallow reliefs are just visible. Moreover, such an unbalanced image is poorly suited to publication in a magazine, which requires pronounced differences in levels between dark and bright areas. A simple increase of the levels in dark areas is feasible with a diffuse selection at the price of increased noise and posterization. Too perfect a balance is not desirable because it could lead to a loss of the three-dimensional appearance of the surface: the Moon really is spherical. This delicate balance adjustment is greatly simplified with some special functions:

1. • Photoshop®: Image → adjustment → Shadow/Highlight (Figure 8.1)

2. • GIMP: colors → Retinex

If our favorite software does not offer a similar function, a diffuse selection (Section 8.2) may greatly help to reasonably increase the brightness: the very lowest levels should remain unaltered (to avoid posterization), while the selected area can be brightened.

Although the effect is understated, such a tool is invaluable to unveil shallow reliefs, domes, wrinkle ridges, and reliefs of less than some hundreds of meters which cannot cast a conspicuous shadow. The trap to avoid is exaggerating the application of the gradient correction, which could lead to an apparently planar lunar surface.

Local contrast with diffuse selection

We saw in Section 7.6 how to improve the readability of an image with global contrast adjustment. However, different areas of the image may contain different local contrasts. For instance, a mare or discrete dorsum has a limited dynamic range, and the contrast cannot be expanded without increasing noise and posterization. On the same image, another feature like a young crater may have a strong dynamic range and the signal/noise ratio is high: the contrast may be expanded without increasing the noise (Figure 8.2). Any contrast adjustment applied to the whole image would cause uneven results.

Unlike the two-layer processing, which acts on the full image with two different processes, we have to apply a process on different parts of the image. Generally no more than two distinct regions are involved, but we can imagine that three or more regions could be processed separately.