Denis Vanneste avait fait une comparaison de différents logiciels
ouvrant les fichiers RAW dont "ProcImg".
En réponse, j'avais dit que j'interrogerai l'auteur du Plug-In. Il a eu
la courtoisie de me répondre ... C'est de l'américain pur jus. Voici ma
question et sa réponse :
Pierre wrote:
Hello Davie,
On original RAW files, color are encoded with 12 to ... bits according
to the camera used. When calling MSDCRAW.DLL in a Delphi program, the
images are converted to 8 bits per color format. I think that to
achieve this, you discard the LSB's. So is there a way to retrieve this
lost information ? Can we use the Gamma, Bright or Color Scale to do
that and how if possible ?
Thank you for your help.
Pierre.
About the 12..10..8 bit conversation:
Here's the deal. Let's say you have a 24bit scanner. At first glance you
might say, that's good enough. Because the ending BMP files are also 24
bit. The problem comes into play where each channel (R,G,B) has a value
from 0-255 (256 different values). 255 is full strength and 0 is no
strength. If during your scan you you reach a portion of the image where
the values are 255,255,255,253,253,253,253,253. Your eye IS able to
distinguish between a 2 level change and thus the image would appear
"solarized" Although you would have to look real close. Now your normal
eye will NOT be able to see a 1 level change. I have provided two bmp
files "green1" and "green2" You can see that green1 is solarized and
green2 appears smooth.
Ok, with that in mind, if you had a scanner that could scan 9 bits per
channel instead of 8, then you would gain the benefit from the rounding
to determine if the output value is 253 or 254. See then the above
example scan might then turn out as 255,255,255,254,254,254,253,253,253.
Thus producing a better image. Now if you are just interested in 24 bit
output, then the 9x9x9bit scanner would be sufficient (27bits) But as
technology advances its easier and cheaper to just create 16x16x16 bit
scanners although you don't get better output than from a 27bit scanner :)
Ok, so now it looks like the more bits the better and so you may want
the 10 bits per channel. Well, if you saw how the RGB grid is layed out
and interpolated, you would quickly realize that even though you have 10
bits per channel, there is still a lot of "error diffusion" occuring.
The only CHIP that does NOT have this problem is the Foveon chip in the
Sigma cameras, they are a one-to-one ratio. NO BAYER GRID, NO
INTERPOLATION. Each cell represents R+G+B values. The only problem is
that the Foveon chip has LOW tolerance. That means in you are in a
studio taking pics with perfect lighting, then your photos will be the
absolute best. If you attempt to up the ISO to 200 or higher, you will
see a DRAMATIC INCREASE in grain. That's because they are working with a
low signal to noise ratio.
Ok, so basically, if you are using Foveon photos, then yeah sure save
the stuff in 16bit per channel PPM, but for all others that use
interpolation you will not gain sharpness AT ALL!!!!!!!!!!!!!!!!!!!!!
Currently there is no provision to extract to 16bit per channel ppm
because these 16 bit files are much much bigger, AND you do NOT see the
photo as it was intended (ie: the photos are super dark) Play with Photo
Shop and you will see what I am saying about 16 bit files. As far as
lopping off the bottom bits, that 's not what happens the data is all
interpolated with the BAYER grid and ONLY AFTER the calculations are
done does the extra bits go away,. This way you get the benefit of
rounding like in the 9x9x9 scanner example I gave you earlier.
Davie
Si j'essaie de résumer (j'espère que ma traduction n'est pas trop
erronée, Davie dit que de toute façon, les interpolations réalisées par
les filtres de BAYER pour recomposer l'image sont telles qu'au delà de 9
bits il n'y a plus rien à récupérer.
D'après lui, seuls les capteurs FOVEON, ne nécessitant pas de tels
filtres sont capables de restituer plus de 9 bits. Ces capteurs ont par
ailleurs d'autres défauts ...
Pour ma part, il me semble qu'en jouant sur les facteurs tels Gamma,
Luminosité et Facteur d'échelle, on peut réhausser les teintes sombres
(les teintes claires sont perdues dans ce cas --> travail avec des
calques) mais Davie dit que ce qu'on ressort, est fortement perturbé par
les traitements liés au filtre BAYER.
Cette action est irreversible, confirmez la suppression du commentaire ?
Signaler le commentaire
Veuillez sélectionner un problème
Nudité
Violence
Harcèlement
Fraude
Vente illégale
Discours haineux
Terrorisme
Autre
Thebe
Pierre CHAUVEAU wrote:
Bonjour à toutes et à tous,
Denis Vanneste avait fait une comparaison de différents logiciels ouvrant les fichiers RAW dont "ProcImg".
En réponse, j'avais dit que j'interrogerai l'auteur du Plug-In. Il a eu la courtoisie de me répondre ... C'est de l'américain pur jus. Voici ma question et sa réponse :
Pierre wrote:
Hello Davie,
On original RAW files, color are encoded with 12 to ... bits according to the camera used. When calling MSDCRAW.DLL in a Delphi program, the images are converted to 8 bits per color format. I think that to achieve this, you discard the LSB's. So is there a way to retrieve this lost information ? Can we use the Gamma, Bright or Color Scale to do that and how if possible ?
Thank you for your help.
Pierre.
About the 12..10..8 bit conversation: Here's the deal. Let's say you have a 24bit scanner. At first glance you might say, that's good enough. Because the ending BMP files are also 24 bit. The problem comes into play where each channel (R,G,B) has a value from 0-255 (256 different values). 255 is full strength and 0 is no strength. If during your scan you you reach a portion of the image where the values are 255,255,255,253,253,253,253,253. Your eye IS able to distinguish between a 2 level change and thus the image would appear "solarized" Although you would have to look real close. Now your normal eye will NOT be able to see a 1 level change. I have provided two bmp files "green1" and "green2" You can see that green1 is solarized and green2 appears smooth.
Ok, with that in mind, if you had a scanner that could scan 9 bits per channel instead of 8, then you would gain the benefit from the rounding to determine if the output value is 253 or 254. See then the above example scan might then turn out as 255,255,255,254,254,254,253,253,253. Thus producing a better image. Now if you are just interested in 24 bit output, then the 9x9x9bit scanner would be sufficient (27bits) But as technology advances its easier and cheaper to just create 16x16x16 bit scanners although you don't get better output than from a 27bit scanner :)
Ok, so now it looks like the more bits the better and so you may want the 10 bits per channel. Well, if you saw how the RGB grid is layed out and interpolated, you would quickly realize that even though you have 10 bits per channel, there is still a lot of "error diffusion" occuring. The only CHIP that does NOT have this problem is the Foveon chip in the Sigma cameras, they are a one-to-one ratio. NO BAYER GRID, NO INTERPOLATION. Each cell represents R+G+B values. The only problem is that the Foveon chip has LOW tolerance. That means in you are in a studio taking pics with perfect lighting, then your photos will be the absolute best. If you attempt to up the ISO to 200 or higher, you will see a DRAMATIC INCREASE in grain. That's because they are working with a low signal to noise ratio.
Ok, so basically, if you are using Foveon photos, then yeah sure save the stuff in 16bit per channel PPM, but for all others that use interpolation you will not gain sharpness AT ALL!!!!!!!!!!!!!!!!!!!!!
Currently there is no provision to extract to 16bit per channel ppm because these 16 bit files are much much bigger, AND you do NOT see the photo as it was intended (ie: the photos are super dark) Play with Photo Shop and you will see what I am saying about 16 bit files. As far as lopping off the bottom bits, that 's not what happens the data is all interpolated with the BAYER grid and ONLY AFTER the calculations are done does the extra bits go away,. This way you get the benefit of rounding like in the 9x9x9 scanner example I gave you earlier.
Davie
Si j'essaie de résumer (j'espère que ma traduction n'est pas trop erronée, Davie dit que de toute façon, les interpolations réalisées par les filtres de BAYER pour recomposer l'image sont telles qu'au delà de 9 bits il n'y a plus rien à récupérer.
D'après lui, seuls les capteurs FOVEON, ne nécessitant pas de tels filtres sont capables de restituer plus de 9 bits. Ces capteurs ont par ailleurs d'autres défauts ...
Pour ma part, il me semble qu'en jouant sur les facteurs tels Gamma, Luminosité et Facteur d'échelle, on peut réhausser les teintes sombres (les teintes claires sont perdues dans ce cas --> travail avec des calques) mais Davie dit que ce qu'on ressort, est fortement perturbé par les traitements liés au filtre BAYER.
Je vais essayer d'en savoir plus.
Pierre.
A ta place, je changerais de logiciel
Pierre CHAUVEAU wrote:
Bonjour à toutes et à tous,
Denis Vanneste avait fait une comparaison de différents logiciels
ouvrant les fichiers RAW dont "ProcImg".
En réponse, j'avais dit que j'interrogerai l'auteur du Plug-In. Il a eu
la courtoisie de me répondre ... C'est de l'américain pur jus. Voici ma
question et sa réponse :
Pierre wrote:
Hello Davie,
On original RAW files, color are encoded with 12 to ... bits according
to the camera used. When calling MSDCRAW.DLL in a Delphi program, the
images are converted to 8 bits per color format. I think that to
achieve this, you discard the LSB's. So is there a way to retrieve this
lost information ? Can we use the Gamma, Bright or Color Scale to do
that and how if possible ?
Thank you for your help.
Pierre.
About the 12..10..8 bit conversation:
Here's the deal. Let's say you have a 24bit scanner. At first glance you
might say, that's good enough. Because the ending BMP files are also 24
bit. The problem comes into play where each channel (R,G,B) has a value
from 0-255 (256 different values). 255 is full strength and 0 is no
strength. If during your scan you you reach a portion of the image where
the values are 255,255,255,253,253,253,253,253. Your eye IS able to
distinguish between a 2 level change and thus the image would appear
"solarized" Although you would have to look real close. Now your normal
eye will NOT be able to see a 1 level change. I have provided two bmp
files "green1" and "green2" You can see that green1 is solarized and
green2 appears smooth.
Ok, with that in mind, if you had a scanner that could scan 9 bits per
channel instead of 8, then you would gain the benefit from the rounding
to determine if the output value is 253 or 254. See then the above
example scan might then turn out as 255,255,255,254,254,254,253,253,253.
Thus producing a better image. Now if you are just interested in 24 bit
output, then the 9x9x9bit scanner would be sufficient (27bits) But as
technology advances its easier and cheaper to just create 16x16x16 bit
scanners although you don't get better output than from a 27bit scanner :)
Ok, so now it looks like the more bits the better and so you may want
the 10 bits per channel. Well, if you saw how the RGB grid is layed out
and interpolated, you would quickly realize that even though you have 10
bits per channel, there is still a lot of "error diffusion" occuring.
The only CHIP that does NOT have this problem is the Foveon chip in the
Sigma cameras, they are a one-to-one ratio. NO BAYER GRID, NO
INTERPOLATION. Each cell represents R+G+B values. The only problem is
that the Foveon chip has LOW tolerance. That means in you are in a
studio taking pics with perfect lighting, then your photos will be the
absolute best. If you attempt to up the ISO to 200 or higher, you will
see a DRAMATIC INCREASE in grain. That's because they are working with a
low signal to noise ratio.
Ok, so basically, if you are using Foveon photos, then yeah sure save
the stuff in 16bit per channel PPM, but for all others that use
interpolation you will not gain sharpness AT ALL!!!!!!!!!!!!!!!!!!!!!
Currently there is no provision to extract to 16bit per channel ppm
because these 16 bit files are much much bigger, AND you do NOT see the
photo as it was intended (ie: the photos are super dark) Play with Photo
Shop and you will see what I am saying about 16 bit files. As far as
lopping off the bottom bits, that 's not what happens the data is all
interpolated with the BAYER grid and ONLY AFTER the calculations are
done does the extra bits go away,. This way you get the benefit of
rounding like in the 9x9x9 scanner example I gave you earlier.
Davie
Si j'essaie de résumer (j'espère que ma traduction n'est pas trop
erronée, Davie dit que de toute façon, les interpolations réalisées par
les filtres de BAYER pour recomposer l'image sont telles qu'au delà de 9
bits il n'y a plus rien à récupérer.
D'après lui, seuls les capteurs FOVEON, ne nécessitant pas de tels
filtres sont capables de restituer plus de 9 bits. Ces capteurs ont par
ailleurs d'autres défauts ...
Pour ma part, il me semble qu'en jouant sur les facteurs tels Gamma,
Luminosité et Facteur d'échelle, on peut réhausser les teintes sombres
(les teintes claires sont perdues dans ce cas --> travail avec des
calques) mais Davie dit que ce qu'on ressort, est fortement perturbé par
les traitements liés au filtre BAYER.
Denis Vanneste avait fait une comparaison de différents logiciels ouvrant les fichiers RAW dont "ProcImg".
En réponse, j'avais dit que j'interrogerai l'auteur du Plug-In. Il a eu la courtoisie de me répondre ... C'est de l'américain pur jus. Voici ma question et sa réponse :
Pierre wrote:
Hello Davie,
On original RAW files, color are encoded with 12 to ... bits according to the camera used. When calling MSDCRAW.DLL in a Delphi program, the images are converted to 8 bits per color format. I think that to achieve this, you discard the LSB's. So is there a way to retrieve this lost information ? Can we use the Gamma, Bright or Color Scale to do that and how if possible ?
Thank you for your help.
Pierre.
About the 12..10..8 bit conversation: Here's the deal. Let's say you have a 24bit scanner. At first glance you might say, that's good enough. Because the ending BMP files are also 24 bit. The problem comes into play where each channel (R,G,B) has a value from 0-255 (256 different values). 255 is full strength and 0 is no strength. If during your scan you you reach a portion of the image where the values are 255,255,255,253,253,253,253,253. Your eye IS able to distinguish between a 2 level change and thus the image would appear "solarized" Although you would have to look real close. Now your normal eye will NOT be able to see a 1 level change. I have provided two bmp files "green1" and "green2" You can see that green1 is solarized and green2 appears smooth.
Ok, with that in mind, if you had a scanner that could scan 9 bits per channel instead of 8, then you would gain the benefit from the rounding to determine if the output value is 253 or 254. See then the above example scan might then turn out as 255,255,255,254,254,254,253,253,253. Thus producing a better image. Now if you are just interested in 24 bit output, then the 9x9x9bit scanner would be sufficient (27bits) But as technology advances its easier and cheaper to just create 16x16x16 bit scanners although you don't get better output than from a 27bit scanner :)
Ok, so now it looks like the more bits the better and so you may want the 10 bits per channel. Well, if you saw how the RGB grid is layed out and interpolated, you would quickly realize that even though you have 10 bits per channel, there is still a lot of "error diffusion" occuring. The only CHIP that does NOT have this problem is the Foveon chip in the Sigma cameras, they are a one-to-one ratio. NO BAYER GRID, NO INTERPOLATION. Each cell represents R+G+B values. The only problem is that the Foveon chip has LOW tolerance. That means in you are in a studio taking pics with perfect lighting, then your photos will be the absolute best. If you attempt to up the ISO to 200 or higher, you will see a DRAMATIC INCREASE in grain. That's because they are working with a low signal to noise ratio.
Ok, so basically, if you are using Foveon photos, then yeah sure save the stuff in 16bit per channel PPM, but for all others that use interpolation you will not gain sharpness AT ALL!!!!!!!!!!!!!!!!!!!!!
Currently there is no provision to extract to 16bit per channel ppm because these 16 bit files are much much bigger, AND you do NOT see the photo as it was intended (ie: the photos are super dark) Play with Photo Shop and you will see what I am saying about 16 bit files. As far as lopping off the bottom bits, that 's not what happens the data is all interpolated with the BAYER grid and ONLY AFTER the calculations are done does the extra bits go away,. This way you get the benefit of rounding like in the 9x9x9 scanner example I gave you earlier.
Davie
Si j'essaie de résumer (j'espère que ma traduction n'est pas trop erronée, Davie dit que de toute façon, les interpolations réalisées par les filtres de BAYER pour recomposer l'image sont telles qu'au delà de 9 bits il n'y a plus rien à récupérer.
D'après lui, seuls les capteurs FOVEON, ne nécessitant pas de tels filtres sont capables de restituer plus de 9 bits. Ces capteurs ont par ailleurs d'autres défauts ...
Pour ma part, il me semble qu'en jouant sur les facteurs tels Gamma, Luminosité et Facteur d'échelle, on peut réhausser les teintes sombres (les teintes claires sont perdues dans ce cas --> travail avec des calques) mais Davie dit que ce qu'on ressort, est fortement perturbé par les traitements liés au filtre BAYER.
Je vais essayer d'en savoir plus.
Pierre.
A ta place, je changerais de logiciel
Pierre Pallier
Hello, Thebe a écrit dans <news:422df16b$0$30303$
Pierre CHAUVEAU wrote:
[104 lignes citées masquées] A ta place, je changerais de logiciel
Je peux te conseiller la même chose ? Ou de brancher le cerveau. -- Pour l'aiguille du trou du sténopé, on trouvera bien une botte de paille. -+- Noëlle sur frp -+-
Hello, Thebe a écrit dans <news:422df16b$0$30303$626a14ce@news.free.fr>
Pierre CHAUVEAU wrote:
[104 lignes citées masquées]
A ta place, je changerais de logiciel
Je peux te conseiller la même chose ?
Ou de brancher le cerveau.
--
Pour l'aiguille du trou du sténopé, on trouvera bien une botte de paille.
-+- Noëlle sur frp -+-
[104 lignes citées masquées] A ta place, je changerais de logiciel
Je peux te conseiller la même chose ? Ou de brancher le cerveau. -- Pour l'aiguille du trou du sténopé, on trouvera bien une botte de paille. -+- Noëlle sur frp -+-
Thebe
Pierre Pallier wrote:
Hello, Thebe a écrit dans <news:422df16b$0$30303$
Pierre CHAUVEAU wrote:
[104 lignes citées masquées]
A ta place, je changerais de logiciel
Je peux te conseiller la même chose ? Ou de brancher le cerveau. Le mien n'est pas fait pour ces niaiseries. De plus je suis convaincu
que les photos sont nulles.
Pierre Pallier wrote:
Hello, Thebe a écrit dans <news:422df16b$0$30303$626a14ce@news.free.fr>
Pierre CHAUVEAU wrote:
[104 lignes citées masquées]
A ta place, je changerais de logiciel
Je peux te conseiller la même chose ?
Ou de brancher le cerveau.
Le mien n'est pas fait pour ces niaiseries. De plus je suis convaincu
