IR and radio-band observations of heavily extinguished regions in starburst galaxies suggest a high supernova (SN) rate associated with such regions. Optically measured SN rates may therefore underestimate the total SN rate by factors of up to 10, as a result of the very high extinction mag) (A B D 10È20 to core-collapse SNe in starburst regions. The IR/radio SN rates come from a variety of indirect means, however, which su †er from model dependence and other problems. We describe a direct measurement of the SN rate from a regular patrol of starburst galaxies done with K@-band imaging to minimize the e †ects of extinction. A collection of K@-band measurements of core-collapse SNe near maximum light is presented. Such measurements (excluding 1987A) are not well reported in the literature. Results of a preliminary K@-band search, using the MIRC camera at the Wyoming Infrared Observatory and an improved search strategy using the new ORCA optics, are described. A monthly patrol of a sample of IRAS bright (mostly starburst) galaxies within 25 Mpc should yield 1È6 SNe yr~1, corresponding to the range of estimated SN rates. Our initial MIRC search with low resolution pixels) failed to Ðnd (2A .2 extinguished SNe in the IRAS galaxies, limiting the SN rate outside the nucleus (at greater than 15A radius) to less than 3.8 far-IR SN rate units (SNe per century per 1010 measured at 60 and 100 km, L _ or FIRSRU) at 90% conÐdence. The MIRC camera had insufficient resolution to search nuclear starburst regions, where starburst and SN activity is concentrated ; therefore, we were unable to rigorously test the hypothesis of high SN rates in heavily obscured star-forming regions. We conclude that highresolution nuclear SN searches in starburst galaxies with small Ðelds are more productive than lowresolution, large-Ðeld searches, even for our sample of large (often several arcminutes) galaxies. With our ORCA high-resolution optics, we could limit the total SN rate to less than 1.3 FIRSRU at 90% conÐdence in 3 years of observations, lower than most estimates.