Why 8x8 block size is preferable in DCT during compression not 4x4 or 64x64 ?

WHy in DCT 8x8 blocks of image are taken and preferred? Why don't we take 4x4, 64x64 blocks wether doing compression using DCT or simple DCT?

Answer 1

[Matt J]

It was probably a result of experimentation/optimization on the part of those who defined the JPEG standard.

If you take the blocks too large, the image will be less smooth over each block, and you will sacrifice compression.

If you make the blocks too small, quantization becomes less flexible. For example, suppose in the extreme that you used 1x1 blocks. Then, the only quantization you can do is to set an entire block to zero, completely destroying it, or leave it as is in which case there is no compression.

[Libra]

1x1 will be very small... I read power of 2 blocks are usually taken then why not 4x4? computations would be much easier and image will be smoother as well?

[Matt J]

1x1 will be very small...I read power of 2 blocks are usually taken then why not 4x4?

4x4 is also very small.