A colour image is really three images, a red one, a green one and a blue one. If you want to do image processing things to a colour image, smoothing for example, then you have to apply your algorithm to the three colours separately. As we have seen, images are stored on computers in various file formats. Just like computer screens, some image formats (tiff, jpeg, ppm, some .png) are 24 bit true colour and store a red, a green and a blue image. Others (.gif, some .png) store an 8-bit image and a colour table to say which value in the image should be what colour.
Try
window,0,retain=2 loadct,39 tvlct,r,g,b,/get tv,bytscl(meteosat) write_png,'~jmarsham/out_39.png',tvrd(),r,g,band
window,0,retain=2 load_sst_ct tvlct,r,g,b,/get tv,bytscl(meteosat) write_png,'~jmarsham/out_sst.png',tvrd(),r,g,b
The different colour tables make a significant difference.