Document QuantizationTableType::Custom format

[sophie-h]

I want to re-encode a JPEG with the same settings it was encoded before. I have the quantization table extracted from the original image in it's either 8bit or 16bit format. But I'm failing to feed it back into jpeg-encoder such that I end up with a new JPEG that uses the same quantization table.

What exactly is the format that QuantizationTableType::Custom expects? Apparently it's not a QT scaled to u16. But u8 data also doesn't give the desired result.

I found this mysterious code

    fn get_user_table(table: &[u16; 64]) -> [NonZeroU16; 64] {
        let mut q_table = [NonZeroU16::new(1).unwrap(); 64];
        for (i, &v) in table.iter().enumerate() {
            q_table[i] = match NonZeroU16::new(v.clamp(1, 2 << 10) << 3) {
                Some(v) => v,
                None => panic!("Invalid quantization table value: {}", v),
            };
        }
        q_table
    }

that seems to limit the input value to 2048 and then left shifts the complete thing by 3. Is that the key to understanding what the input format is?

[sophie-h]

I noticed that the matrices still contain the same numbers before and after. They are just re-arranged. Does QuantizationTableType::Custom expect a different order than is used in the DQT segment?

1	1	1	1	1	1	1	1
1	1	1	1	1	1	1	2
1	1	1	1	1	2	2	2
1	2	3	3	3	3	3	3
3	3	3	4	5	4	3	3
4	3	3	3	4	5	4	4
5	5	5	5	5	3	4	6
6	6	5	6	5	5	5	5

1	1	1	1	1	1	1	1
1	1	3	2	1	1	1	1
1	1	3	3	4	5	3	3
3	1	1	1	1	1	2	3
4	3	5	6	6	5	3	5
3	2	2	2	3	4	4	5
5	6	5	5	3	3	3	4
3	5	5	4	4	6	5	5

[sophie-h]

Okay, looks like the input is in 8-bit and not in zig-zag sequence