Neither. The Sahara Desert is heterogenous. It’s yellow, red, orange, brown, black, and even green in certain places and at certain times of year.
Slightly less cheeky answer: neither. Both examples are both approximations of human vision, displayed on computer screens that can be highly variable. The 2002 version of the Blue Marble is a composite of 4 months of atmospherically-corrected data in narrow wavelength bands. The composite data is scaled to approximate the nonlinear response of human vision. In particularly bright areas the image is clipped in some bands, leading to incorrect hue and loss of contrast.
For all the gory details, here’s the user’s guide for the MODIS land surface reflectance product (which has always looked slightly greenish to me): http://modis-sr.ltdri.org/guide/MOD09_UserGuide_v1.4.pdf
The Apollo 17 Blue Marble is a color-corrected scan of 40-year-old film (stored in a freezer at NASA’s Johnson Space Center) and is also not a particularly good representation of what an astronaut would see (as Pont commented earlier). The best description of the appearance of the Earth from space is from astronaut Piers Sellers, who said it’s “brilliant blue, and too bright to look at directly—like sunlight reflected off snow while skiing at high altitude”.
Brutally honest answer: neither. To paraphrase Mapbox’s Charlie Loyd, there’s an irreducible subjectivity to color. How we perceive color is affected by ambient light, adjacent colors, overall amount of light, individual physiology, our age, and probably mood and what we had to eat for breakfast. Viewed from space there are even more variables: particularly atmospheric conditions (which influence both incident and reflected light) and viewing angle.
Current state-of-the-art is probably NOAA View, http://www.nnvl.noaa.gov/view/#TRUE which uses a multi-band algorithm to simulate true color with Visible Infrared Imaging Radiometer Suite (VIIRS) data. According to that algorithm, the Sahara is more red than yellow.
For the record, I’ve worked on color-correcting MODIS imagery, the Blue Marble datasets, and astronaut photography on the NASA Earth Observatory & Visible Earth sites (including the referenced version of AS17-148-22727—unfortunately I can’t find the original TIFF.) Perhaps that means I think the Sahara is both ochre and yellow. I also wrote the guide to color-correcting Landsat imagery linked by Jezibelle (thank you).