Checked textile designs have many repeated patterns, so their distributions in the spatial frequency domains are partially concentrated. In addition, each distribution around local values in the spatial frequency domains can be approximately regarded as the normal distribution. By this property, it is possible to reconstruct distributions close to original distributions by using characteristic values of each distribution around local values. Moreover, in checked images, it is possible to obtain reconstructed images close to original images by using the data on the vertical and horizontal axes in the spatial frequency domains. So, we focused on reconstructing only distributions on the vertical and horizontal axes in the spatial frequency domains to compress checked images and used Gram-Charlier approximate equation to reconstruct the distributions. In this method, the mean, valiance, skewness and kurtosis [1-3] were used as characteristic values of the distributive shapes. Finally, we used only local values and these four data of distributive shapes around the local values as the compressed codes, and the other data aren't required. We propose the compressing method using these properties and show the effectiveness for checked images.