This paper presents a color transformation algorithm into the preferred colors extracted from the reference image. Key color objects areas are separated from original image by the image segmentation method using color distance measures in CIELAB space and the filtering considering local information. Each object color is transformed for its principal component (PC) to be matched to the corresponding reference color what we feel to be preferable. The color conversion into preferred reference color is performed by 3 x 3 matrix derived by PCA. Finally the preferred image is obtained by merging these key color areas into other remaining areas. The paper introduces the experimental results on the transformation for memorial colors such as skin color, green grass, and blue sky. extracted from different reference image and discusses the reduction of segmentation errors.