Other posters have pointed out that it's the hazard avoidance camera, but they haven't said why the hazard avoidance camera is black and white.
When you do computer vision, the first step you do is convert your color image into a black and white image, and run your CV algorithms on the black and white image. This is because when you're looking at objects and shapes and stuff, it's contrast that tells you where the boundaries between things are. This is true even in a human world of human objects, which tend to be many colored. It's even more true on Mars where basically everything is varying shades of orange. So having color doesn't help a whole lot, and you also have to do the additional step of converting the color image to black and white, which takes CPU power and adds latency. Remember, the purpose is hazard avoidance- latency is bad.
Additionally, color camera sensors aren't actually color sensors. They're black and white sensors. In front of every pixel on the black and white sensor is a filter that is either red, green, or blue. Pixels are grouped into sets of four, and there are two pixels with green filters, one pixel with a blue filter, and one filter with a red filter. (sometimes one of the green filters is omitted, giving red, green, blue, and b&w, or sometimes one of the green filters is a filter that allows IR, or something like that.) So if you have a 16MP camera, the camera has 8M green, 4M red, and 4M blue pixels. This means two things; first of all, if you just wanted a black and white image in the first place, a color sensor gives less detail than the equivalent black and white sensor, and second, you need to do additional processing to convert the raw output from the sensor into an image that's usable for anything. The additional processing adds latency.