In this work, some sea viewing wide field-of-view sensor (SeaWiFS) images over the Santa Barbara Channel (SBC) are analysed. Pixel-by-pixel measurements of radiances at eight SeaWiFS channels and analytic Jacobians are simulated using a coupled atmosphere-ocean radiative transfer model. The inverse algorithm is based on optimal estimation with loosely constrained a priori data. The five-element state vector has two aerosol (optical depth at 865 nm, bimodal fraction of particles) and three marine (chlorophyll concentration, detrital/dissolved-matter absorption at 443 nm, and backscattering coefficient at 443 nm) parameters. The retrieval is stable and well posed; the results are smoother and show less spread than those derived from the standard SeaDAS v4.8 algorithm. For a 28 February 2003 SeaWiFS image, the average radiance residual is less than 1% for seven SeaWiFS channels, and less than 2% for the 765 nm channel. For a series of SBC SeaWiFS match-up cases over a 4-year period, estimated water-leaving radiances agree well with field measurements.