Cloud motion vectors derived from sequences of remotely sensed data are widely used by numerical weather prediction models and other meteorological and climatic applications. One approach to computing cloud motion vectors is the correlation-relaxation labeling technique, in which a set of candidate vectors for each template is refined using relaxation labeling to provide a local smoothness constraint. In this letter, an extension of the correlation-relaxation labeling framework to tracking clouds in multichannel imagery is presented. As this multichannel approach takes advantage of the diversity between channels, it has the potential for producing motion vectors with a superior quality and coverage than can be achieved by any individual channel. Results for visible and infrared images from Meteostat Second Generation confirm the benefits of the multichannel approach.