A spatial gamut mapping technique is proposed to overcome the shortcomings encountered with standard pointwise gamut mapping algorithms by preserving spatially local luminance variations in the original image. It does so by first processing the image through a standard pointwise gamut mapping algorithm. The difference between the original image luminance Y and gamut mapped image luminance Y' is calculated. A spatial filter is then applied to this difference signal, and added back to the gamut mapped signal Y'. The filtering operation can result in colors near the gamut boundary being placed outside the gamut, hence a second gamut mapping step is required to move these pixels back into the gamut. Finally, the in-gamut pixels are processed through a color correction function for the output device, and rendered to that device. Psychophysical experiments validate the superior performance of the proposed algorithm, which reduces many of the artifacts arising from standard pointwise techniques.