PNG or JPEG for an X-Ray without visible artifacts

Question

A past paper question from a Multimedia Systems exam:

If you had an X-Ray photograph of a leg which might be broken, which would give better compression before artifacts were visible, and which would you choose?

I think the answer is JPEG. However, my class mate thinks it is PNG.

I chose JPEG because:

1. The JPEG could be set to a low compression ratio (such that the quantization matrix will preserve most of the high frequencies from the DCT and still reduce the size of the coefficients). However, as a result, run-length encoding would yield significantly lower lossless compression ratio because there will be fewer zeroed high frequency coefficients, and as a result there many not be any subset of values with significantly large frequencies.

2. PNG is exlusively lossless compression which uses DEFLATE (LZ77 with Huffman). However, even though a X-Ray is mostly black and white, there are very subtle differences in the luminosity between pixels which would mean that PNG would not be able to achieve a high compression ratio. The PNG would be significantly larger than the JPEG before the compression on the JPEG was such that it would yield visible artifacts.

I'm looking for an authoritiatve answer. Is my reasoning sound, or would PNG be the better image format?

Answer 1

It's the JPEG. The keyword here is visible. The entire reason for JPEG working so well is that you can quantize away lots of high frequencies before the human visual system can even perceive any artifacts.

Also note that before artifacts (like ringing) become present, the picture will simply lose high frequencies. Without a reference picture to compare against, a human would probably not see the absence of high frequencies.

Your other assumptions are all correct.