Abstract
A new image coding method, JPEG 2000, has a structure that allows separate a input image to sub-image 'tiles' and some discrete wavelet transform can apply to each tiles. In recently standardized JPEG 2000 Part 1, a number of specified transforms is only two. One is for irreversible coding, and another is for reversible coding. Though, in currently standardizing JPEG 2000 Part 2, any tranforms can be specified for each tiles, and this enables adaptive coding on tiles. In this report, what tile size is efficient for this adaptive coding is reported.
