This paper proposes multi-level vector error diffusion based on adaptive primary color selection for fast and accurate color reproduction. Conventional bi-level vector error diffusion uses eight primary colors(R, G, B, C, M, Y, W, K). However, multi-level vector error diffusion uses more primary colors (this paper uses 64 primary colors) depending on the printing device, thereby significantly increasing the time complexity due to the additional increment of computation. Moreover, the output image can also include color artifacts that have a noticeable primary color under the influence of
large quantization error and increased primary color. Accordingly, to reduce these problems, we proposed the quantization process to decide a candidate primary among the 64 primary colors using lightness difference. First, we classified the 64 primary colors into 60 chromatic colors and 4 achromatic colors and then we exclude primary colors with the large lightness difference against the input color from a set of 60 chromatic primary colors. Using both 4 achromatic primary colors and a candidate primary colors, we calculated the vector norm to select output color. Also this paper determine optimal threshold experimentally to remove smear artifacts resulting from the diffusion of large quantization error. As a result, this paper archives fast multi-level vector error diffusion by avoiding additional computation and produces visually pleasing halftone pattern by excluding noticeable primary colors.