Random grid (RG) is an approach to implement visual secret sharing (VSS) without pixel expansion. However, visual quality of the recovered secret image in RG-based VSS is not satisfactory. In this paper, two methodologies are introduced for improving image quality. Firstly, a random noise balanced error diffusion (RNBED) algorithm is proposed for generating RGs whose black pixels are distributed homogeneously. By combining the proposed RNBED algorithm and existing RG-based VSS schemes, two approaches for enhancing the recovered image quality are presented. Experimental results are provided, illustrating that competitive visual quality is achieved.