To investigate the accuracy of a process of colour reproduction of water colour images based on spectral data, whereby the reproduction images were presented on LCD. In the process, a mono-chromatic digital video camera with 640x480 pixels was used to capture the original water colour images through 5 optimized filters: BPB42, BPB53, BPN50, SC64 and SC66. Principal component analysis (PCA) was employed to determine the optimum number of filters. The types of filters to be used were determined by the minimum colour difference ([delta]E*[subscript ab]) between the original and reproduction of the training set reproduced through different types of filters. Spectral data of the images were estimated using Wiener estimation method, then converted to XYZ values and sRGB values, respectively. Before displaying the reproduction images, the colour information was transformed to RGB data using colour management software, so that the LCD would produce the accurate colour images. Finally, the originals were compared quantitatively and subjectively to the reproductions under D[subscript 50] and D[subscript 65]. From the result of quantitative analysis, it was found that the average [[delta]E*[subscript ab] of skin tone, leave tone and sky tone were 8.74, 10.15 and 6.88, respectively, under D[subscript 50] and were 8.88, 10.16 and 7.02, respectively, under D[subscript 65]. In the case of subjective analysis, the reproduction images displayed on the LCD were similar to the original water colour images only for skin tone under D[subscript 50] with medium level and yellow tone under D[subscript 50] and D[subscript 65] with high level.