We have been proposed a GBD (Gamut Boundary Descriptor) for comparing the gamut between image and device in discrete polar angle segment divided by (Δ, Δφ) in CIELAB space. The gamut shell is described as the simple radial distances, called r-image. This GBD represents a monochromatic 2D image whose pixel denotes the radial vector magnitude arranged in discrete integer address. This simple presentation makes it easy to compare the point-to-point gamut sizes between image and device. We presented a method of image-dependent gamut compression using r-image and introduced image-dependent gamut expansion using histogram stretch. This paper examines how to decide which should be applied, compression or expansion, for gamut mapping in reference to the each r-image of image and device. When the most of r-image pixel values for image are greater than those for device in entire region and around middle lightness region, a gamut compression is necessary. On the other hand, if they are smaller than those for device in middle region and mainly located at lower region of lightness, a gamut expansion is desirable. The numerical gamut comparison by r-image to decide compression or expansion is tested for typical image samples. The paper reports the experimental results in both compression for wide gamut and expansion for unsaturated narrow gamut.