JPEG compression works in successive layers. In DCT Layer, which compresses each component (Y,Cb,Cr) individually it will consider 8x8 image blocks because it gets easier to find those details which can be removed so as to get image with less file size without loosing information (aim of compression). If we apply it to whole image then changes are quite visible. Eg. If you apply DCT to Y component of entire image you will get image with less brightness.