> >> For me a compressed image format is a format that have losses. e.g.
> >> I take 5*5 pixels and calcul the mean of it. So the size is smaller
> >> as further. In the normal status you dont see the quality losses
> >> but if you zoom in. Do you understand, what I mean with compressed?
> >> OK but after I had saved the images I want to analyse it with
> >> Matlab. And I've been afraid that if I use compressed images which
> >> had information losses, the results are wrong. That's all.
> >
> >
> > An entirely valid reason to avoid any lossy image format.
> >
> Absolutely. But then the first thing to avoid is to go through YUV420P
> (which uses only 1.5 bits of info per pixel) or even through any other
> pixel format than the one that will be used for the image processing in
> Matlab.
>
> So, the specs are: 25fps (BTW at what size and for how long a time), a
> lossless codec, which has to be acceptable to MATLAB, and no lossy
> colorspace conversions. Thus JPEG is out (lossy), and XPM too (uses a
> colormap). That leaves BMP and maybe GIF (does a TrueColr GIF exist?),
> and maybe PGM and PPM - I dont know much about them, are they really
> lossless *and* free of colorspace conversions?
>
You lose color information using YUV420 (and it's 1.5 bytes/pixel.) (*)
Only the brightness part of the image is unchanged. PPM is a raw format,
basically same as (non-RLE) 24-bit BMP. But there's no point in
converting YUV420 to RGB24; grab packed RGB24 frames and you got
yourself a "rawbits" PPM minus the header. "man ppm" gives the details.
I have never used Matlab, but my 3rd eye (not the one in the back ;)
tells me that you can read raw binary data into it? In this case one
could use those raw buffers as they are.
Jonk
(*) Provided a true RGB camera is used. With an ordinary PAL camera
no information is lost.
