An image data compression method includes the steps of dividing image data into predetermined regions, generating modeled data by extracting a feature of the image data for each of the regions, performing trial calculation of data quantity of compressed data that is generated by a thinning-out process in which data is thinned out from the modeled data based on a threshold value concerning the thinning-out process, determining a thinning-out threshold value that is a threshold value such that the data quantity becomes equal to or less than a predetermined target data quantity, and performing the thinning-out process based on the thinning-out threshold value so that compressed data is generated.