Accurate sea surface temperatures (SST) are relevant for the work of oceanographers investigating many aspects of the ocean's surface. Satellite imagery provides access to this type of data to a degree not previously attainable. Cloud contamination represents an obstacle to the utilization of the satellite-derived SSTs because it interferes with the retrieval of the temperatures below. The Artificial Neural Network is capable of recognizing patterns. Furthermore, types of neural nets can learn any continuous mapping to an arbitrary accuracy. We investigate the use of one such network, a multi-layer perceptron, for cloud detection of Advanced Very High Resolution Radiometry (AVHRR) SST imagery utilizing the multiple channels contained therein. We find that this approach is suitable for and provides a fast and powerful solution to the problem of detection of cloud-contaminated pixels in satellite imagery.