Image processing reveals features on surface of star

Betelgeuse, the second brightest star in the constellation Orion, is a red supergiant, over 600 times larger than the sun. Although at a distance of approximately 500 light years, viewed from Earth, it has a large apparent angular size, which makes it a prime target for observing the surface activity of stars. However, to observe features on the surface of Betelgeuse, the resolving power of the telescope used must be extremely high. That of ground-based telescopes with an aperture of 10 meters would theoretically be sufficient, but distortion by the Earth's atmosphere greatly limits resolution. Nevertheless, several attempts have been made to observe surface features of Betelgeuse using ground-based techniques.^1–3 Computer processing of the data showed asymmetric structures.

The Hubble Space Telescope (HST) has an aperture of only 2.4 meters, but because it orbits the Earth above the atmosphere it is not affected by distortion. Gilliland and Dupree obtained the first direct image of Betelgeuse with the HST in 1995. The ultraviolet images showed a small bright area, referred to as a hotspot.⁴ These images, however, were not fine enough to reveal the forms of the hotspots. So, to produce images with resolutions comparable to those that would be obtained with a telescope of much larger aperture, we used an image-processing technique called superresolution on data captured by the HST.

Superresolution can produce images with much higher resolution than would be expected from the theoretical limits for a given telescope. We used the three methods recently proposed by Boukouvala and Lettington (BL),⁵ Lucy,⁶ and Miura and Baba (MB)⁷ to produced images with resolutions equivalent to that of telescopes with much larger apertures. From the resulting images, we assessed the relative ability of these methods to extract information about hotspots.

Figure 1 shows images of Betelgeuse in the ultraviolet taken by the Faint Object Camera of the HST from 1995 to 1999. The overall form of the image of Betelgeuse clearly appears to be changing over this period. However, fine features are not discernable because of the low resolving power. We sequentially applied the BL and MB methods to these images to obtain the images in Figure 2. As expected, the ability to discern hotspots was improved.

We first tested the individual performances of each of the three methods.⁸ With the BL method, the superresolution effect was too great, resulting in the loss of disk components and the appearance of undesirable artifacts. In contrast, images obtained with the MB method showed fewer artifacts, but the superresolution effect was insufficient. The Lucy method was also useful for reducing artifacts. We then tried combinations of two superresolution methods. All of the combinations together produced further improved images because one method compensated for the drawbacks of the other. The results of one of these combinations are shown in Figure 2.

We determined the number of hotspots and their positions from the superresolution images and fitted the surface features to model profiles. Figure 3 schematically shows surface features of Betelgeuse. The blue ring indicates the limb-darkened edge of the stellar disk, while red, orange, and yellow rings correspond to the half-width at half-maximum of a Gaussian profile of the hotspot. It is possible to visualize the changes in surface features over time. However, these results are preliminary, and we are now developing better methods for model fitting.

We used combinations of superresolution image-processing techniques to produce images of the star Betelgeuse in the ultraviolet, with higher resolution than would be expected from the theoretical limits of the HST. As a result, we obtained finer details of surface features of the star clearly showing the changing face of Betelgeuse over the period of observation. We expect that our technique will allow the resolution of stellar disks of other stars with apparently large angular sizes.