Anfonwch hwn fel neges destun: Learning texton models for real-time scene context