In message , Mayayana
writes:
"J. P. Gilliver (John)" wrote

| That's true for lossless. But the cropping itself is always destructive.
|
| Other than that cropping obviously removes information, what do you
| mean: I thought the non-destructive crop was just that (in the part of
| the image you keep, obviously). Being as it (as implemented in
| IrfanView, anyway) crops to the nearest 16 (I think it's 16) pixel
| boundary. I assumed the reason it does that is t avoid loss.
|

It's a clever method, but in general editing JPG
is lossy. How often will one need to crop to the
nearest 16 pixels but have no reason to do other
editing? If one will do other editing then the image
should be taken out of JPG format. So it's a kind of
silk purse from a sow's ear thing.

Often, an image is only available as a .JPG: it might have been received
in an email as such, or downloaded from a website; or, the majority of
cameras other than those sold for the serious professional (with
appropriate pricetags) do _not_ offer raw bitmap formats. (They
sometimes offer three _quality_ levels.)

Nospam was just arguing, splitting hairs. It's
really all he does.

Well, we all - including you and I, definitely - dislike imprecision,
especially if it actually results in an untruth being stated (even if
unintentionally). What level of simplification versus imprecision is
acceptable, varies from person to person and between situations: in
other words, one man's desire for precision is another man's
hair-splitting. I can't say I've registered nospam as _particularly_
irritating in that respect, but that's mainly because I tend not to
remember people's levels unless they're _particularly_ irritating (in
which case I'm likely to killfile them, and haven't with him yet), so
you _may_ be right.

| IMO, BMP should only be used when a software doesn't support a better
image
| format. How it stores 24bpp image pixels is unacceptably wasteful.

Re-reading that, I do agree it's oversimplifying - there _is_ no better
format _in terms of accuracy_.
|
| In what way - does it use two 16-bit words, or something? Or do you just
| mean it doesn't do any (even lossless) data-compression?

It has no compression. It's very straightforward.
In general a BMP will be a 24-bit, uncompressed

I was giving the benefit of the doubt: I thought he might have meant
he'd found a case where it used 4 bytes to store the 3 byte information,
or something. If he just means it does no lossy compression, I'd agree
with you; if he means it does no loss_less_ compression, then he should
have made it clearer that that' what he was referring to.

image. (There are other options, but they're no
longer used as far as I know.) The header looks
[]
I _think_ the two-level (one _bit_ per pixel) form is still supported
(e. g. by IrfanView), though I'm not sure if it includes a palette for
the two colours. (I _think_ GIF does have such [and 2- and 4-bit - 4 and
16 colour - modes.)

That's what all raster images are. Pixel grids. Bitmaps.
[Long section snipped - I presume written for readers other than me.]
They're all just ways to store a BMP. None of those
image formats means anything until the BMP is extracted.
One can't render a JPG onscreen any more than the words
of a ZIPped Word DOC can be read from the ZIP bytes.

Good analogy! (Though I'd have probably said text file.)

It has to be decompressed to get the BMP.

Similarly, when one applies filters in an editing program
it's just a math formula applied to the bitmap bytes.
Sharpening increases the difference between the numeric
values. Interpolation for resizing calculates a new pixel
grid by examining the values of neighboring pixels.
Lightening increases the byte values of the pixel bytes.
It's all just math operations on 3-byte RGB pixels stored
as grids in a BMP.

In other words, the idea that BMP is outdated is a
misunderstanding of what raster images are.

Well, I'm not sure if PNG can be lossless. (Actually, I'm not sure if
JP[E]G can; I know the quality slider in IrfanView can be pushed up to
100%, but I think that still involves some loss.)

Then, of course, there are vector images (like good old HPGL, as well as
more modern ones) - let alone fractals! But for actual pictures taken
with a camera, they're all going to be bitmap rasters in the first place
anyway.

[Actually, use of the word raster reminds me: true rendering of
*archive* _video_ material (i. e. shot with a CRT camera) ought to
involve a _slightly_ slanted raster - which, I think, no modern
rendering does.]
--
J. P. Gilliver. UMRA: 1960/1985 MB++G()AL-IS-Ch++(p)Ar@T+H+Sh0!:`)DNAf

And every day in Britain, 33 properties are sold for around that price [a
million pounds or so]. - Jane Rackham, RT 2015/4/11-17