Re: [ML] [550D] False colors
Ba...@aol.com
Here's my first (very rough and slow) port of false colors on 550D.
It's inspired from AJ's 5D2 code, from which I've used the lookup
table.
Quote from AJ:
+) False colours may need 'callibrating'. I have a lookup table for each of
the 256 levels that I've guessed by analysing the Marshal webstie, and
guessing the correlation between the canon 0..255 and the 'IRE' from
Marshall.
Many thanks to AJ, it's a pleasure reading his code (at least the
non-ASM part) :D
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
To unsubscribe from this group, send email to ml-devel+u...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/ml-devel?hl=en
Alex
percent (0-100) and IRE? Is it the same thing?
Ba...@aol.com
Thanks. Do you know what is the transformation between spotmeter's
percent (0-100) and IRE? Is it the same thing?
Alex
Now I'm trying to port AJ's ASM code, in order to make it really fast.
Ba...@aol.com
Nice, it's the same color palette used by AJ.
Now I'm trying to port AJ's ASM code, in order to make it really fast.
Alex
there, you can select it like this:
False Color:
- Plain C: my code, slow
- ASM (AJ): optimized code
Benchmarks (with all other stuff turned off, 720x480 area):
- Plain C: 5-6fps. JaKob got 7 fps.
- ASM (AJ): 14fps.
AJ bug:
- If I toggle Play->LiveView, on first toggle it works, on second
toggle it doesn't work. It alternates. No error message given, no idea
why it happens.
Usability:
- Plain C version does not overwrite Canon displays and other ML
displays, and disappears quickly on half-shutter press.
- ASM version overwrites everything, and doesn't disappear when
stopped. This is due to my strange "mirroring" algorithm (which isn't
used by AJ). I've added a 50px border on every side, and this also
increased the fps to 20.
- None of them can work with focus peaking at the same time.
- Would it be better to trigger this with a shortcut key?
Alex
JaKoB
Matt Dennie
http://img706.imageshack.us/img706/8259/falsecolors.jpg
On Sat, Jan 22, 2011 at 3:08 PM, JaKoB <jako...@gmail.com> wrote:
> Would it be better to trigger this with a shortcut key?
>
> YES YES please
>
Antony Newman
Antony Newman
"AJ bug:
- If I toggle Play->LiveView, on first toggle it works, on second
toggle it doesn't work. It alternates. No error message given, no idea
why it happens"
I think (guess) that when play is pressed, the vram for LiveView is freed, and MoviePlayer malloc new dma memory.
When you go back into LiveView ... the HD video ram is realllocated ... but in a different place ... so you get to magnify something completely random.
AJ
K.
Fernando Freire
I consider usable both, even the 5-7 fps C version.
Some thoughts. This task is CPU intensive, the battery life could be an issue (as it seems on 5D development) and for longer battery life I propose:
Cut borders even more (?), also in C version, to get the size of ML menu (?)
Alex
> Is the bug on a 5dmk2 or the 550D?
On the 550D, of course. Are you reading anything from some DryOs
structures in the code I've ported?
( Changeset: https://bitbucket.org/hudson/magic-lantern/changeset/58618c491f41 )
Border: yes, only this will get around 10fps on the C version
Downsampling (2x2): it will get around 20fps.. I think. My C zebras
(30fps due to msleeps, 40fps with msleep(1)) use 4(x) * 2(y)
downsampling.
Shortcut key: I'll try to hook Dof preview or Flash button
Fernando Freire
Antony Newman
Just a small note - the ASM routine that you are calling:
aj_FalseColour_HistoCalc()
1) If you want a 'quick win' speed up - remove the following bits of code (ie the bits that update the histogram counters, and optionally store them back)
" CMP r4,#1\n" // r5 = False Col + Hist value
" ADDEQ r5,r5,#1\n" // r5 = Hist value
" STREQ r5,[r0, r14]\n" // Write back new Hist value
That's three clock cylces saved per pixel.
2) The other routines (fast histogram) only process alternate vram pixels (about 35% faster)
3) Beyond that I haven't attempt to optimise it ... it can actually be coded run twice as fast - but the code will be 8 times larger.
You'll need to find someone who wants to play with LDM 4 registers at the time....
4) I started coding a version that is 6 times faster - but it is 64 times larger code size (trick, there is no lookup table).
Battery power. I'll leave you guys to decide your best course of action. I've already coded for a power throttle in my code base.
AJ
Alex
- False color area is smaller and faster; also a downsampled (low-res)
version is available (and used by default)
- Shortcut key: Flash button in movie mode, DOF preview in photo mode
(enabled by default, Q to disable)
Other changes:
- Fix from Deti: "This patch fixes focus peaking with HDMI display in
record mode."
- Cropmarks available in Playback mode (use Q to enable this)
- IRE spotmeter (from AJ's excel)
Also, Sawomedia created a page with many cropmarks:
http://www.sawomedia.de/magiclantern/cropmarks/
K.
Antony Newman
Here's a 'histogram free' version for you (edited in gmail .. so there's a chance I've made a mistake!)
You'll need to remove the 'Logical OR' in the calling of it (to remove g_display_histogram)
AJ
/****************************************************************************
* aj_FalseColour() * * *
* False Colour - Use Marshal like false colour *
* http://www.lcdracks.com/monitors/v-lcd70p-hdmi.html *
* Histogram - Find the number of occurances of each range of luma *
* *
* *
* Params: R0 = g_arr_falsColhist .
* R1 = V_start (Std vram) *
* R2 = Vram_pixels (Must be a multiple of 4, and >0)
* R3 = Bmp_adr (Address of first pixel of overlay) *
* *
* Returns: 1 = Success *
* 2 = Failure -> V_start Must be word aligned, and Not Zero *
* -> Bmp Must be word aligned and Not Zero *
* -> Vram_pixels Must be multiple of 4, & not Zero *
* 666 = Voodoo is at work. This should not be possible. *
* *
****************************************************************************/
extern unsigned int aj_FalseColour(
unsigned int g_arr_falsColhist, // R0 = g_arr_falsColhist .
unsigned int V_start, // R1 = first word of Vram to process
unsigned int Vram_pixels, // R2 = process this number of vram pixels
unsigned int Bmp ) // R3 = first word of Overlay to write to
{
// g_arr_falsColhist[256]. Top byte has false Overlay colour for that Luma level
/******
IRE
101 > [254] RED
95 - 100 ORANGE
85 - 94 BRIGHT YELLOW
77 - 84 GREENY YELLOW
59 - 76 LIGHT GREY
54 - 58 PINK
48 - 53 SLIGHTLY DARK GREY
42 - 47 GREEN
22 - 41 DARK GREY
12 - 21 LIGHT BLUE
00 - 11 DARK BLUE
<0 [02] FUCHSIA
******/
//--------------------------------------------------------------------
//------ R E G I S T E R L I S T --------------------------------
//--------------------------------------------------------------------
// R0 = g_arr_falsColhist R7 = Pix1 Vram line1 word1
// R1 = V_start R8 = Pix2 Vram line1 word2
// R2 = Vram_pixels R9 = <NOT USED>
// R3 = Bmp_adr R10 = <NOT USED>
// R4 = <NOT USED> R11 = <NOT USED>
// R5 = Histo_luma_value R12 = <NOT USED>
// R6 = Colour word to write to Overlay R13 = STACK POINTER
// R14 = TEMP WORKSPACE
//--------------------------------------------------------------------
asm volatile(
//=========================================
// SAVE REGISTERS BEFORE WE 'CLOBBER' THEM
//=========================================
" STMFD r13!,{r6,r7,r8,r14}\n" // Save registers that we are going to trash
" BICS r1,#0x3\n" // V_start Must be word aligned, and Not Zero
" BICNES r3,#0x3\n" // Bmp Must be word aligned and Not Zero
" BICNES r2,#0x3\n" // Vram_pixels Must be a multiple of 4, and not Zero
" STMFD r13!,{r5}\n" // Save registers that we are going to trash
" BEQ FLS_RoutineFailure\n" // Return if problem
//+=============================================+
//| MAIN LOOP - Read_4_Vram_Pixels |
//+=============================================+
"FLS_AJ_MAIN_LOOP:\n"
" MOV r6,#0x0\n" // Set Overlay pixel colour = completely see through
" LDMIA r1!,{r7, r8}\n" // LDMIA V_start,{Pix2, Pix3}
//============================================================================================
// Pix1a
//============================================================================================
" AND r14,r7,#0x0000FF00\n" // Luma of Vpixel 1 (shifted)
" MOV r14,r14,LSR#6\n" // Luma x 4 (ie create word position to add)
" LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
" ORR r6,r6,r5,LSR#24\n" // add to bottom (1st) byte of Overlay colour
//============================================================================================
// Pix1b
//============================================================================================
" AND r14,r7,#0xFF000000\n" //
" MOV r14,r14,LSR#24-2\n" // Luma x 4 (ie create word position to add)
" LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
" AND r5,r5,#0xFF<<24\n" // r5 = top byte = False colour -> get it
" ORR r6,r6,r5,LSR#16\n" // add to 2nd byte of Overlay colour
//============================================================================================
// Pix2a
//============================================================================================
" AND r14,r8,#0x0000FF00\n" //
" MOV r14,r14,LSR#6\n" // Luma x 4 (ie create word position to add)
" LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
" AND r5,r5,#0xFF<<24\n" // r5 = top byte = False colour -> get it
" ORR r6,r6,r5,LSR#8\n" // add to bottom (3rd) byte of Overlay colour
//============================================================================================
// Pix2b
//============================================================================================
" AND r14,r8,#0xFF000000\n" //
" MOV r14,r14,LSR#24-2\n" // Luma x 4 (ie create word position to add)
" LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
" AND r5,r5,#0xFF<<24\n" // r5 = top byte = False colour -> get it
" ORR r6,r6,r5\n" // add to 4th byte of Overlay colour
//===============================================
// WriteWordToOverlay
//===============================================
"FLS_WriteWordToOverlay:\n"
" STMIA r3!,{r6}\n" //
//===============================================
// ReadyForNextBlock
//===============================================
" SUBS r2,r2,#4\n" //
" BNE FLS_AJ_MAIN_LOOP\n" //
//===============================================
// RoutineCompleted SUCCESSFULLY
//===============================================
"FLS_RoutineComplete:\n"
" LDMFD r13!,{r5}\n"
" MOV r0,#1\n" // SUCCESS
" LDMFD r13!,{r6,r7,r8,r15}\n"
//===============================================
// RoutineCompleted FAILED
//===============================================
"FLS_RoutineFailure:\n"
" LDMFD r13!,{r5}\n"
" MOV r0,#2\n" // FAILURE
" LDMFD r13!,{r6,r7,r8,r15}\n"
//===============================================
//===============================================
//======= ^^ RETURN POINT OF ASM ^^ ==========
//===============================================
//===============================================
: // Output operands eg. [output1]"+r"(g_r13_temp_store)
: // Input operands eg. [input]"m"(parm1)
: "r0","r1","r2","r3" // eg "memory","cc" = Clobber list
); // end of asm volatile()
return( 666 );
} /* end of aj_FalseColour() */
cristian paradiso
Da: Alex <broscu...@gmail.com>
A: ml-d...@googlegroups.com
Inviato: Dom 23 gennaio 2011, 16:22:00
Oggetto: Re: [ML] [550D] False colors
Alex
algorithm, to avoid overwriting other stuff (here, the focus
rectangle, the spotmeter and part of histogram; maybe the Canon side
menus, if I'll expand the area).
That's the main reason for not using it by default. The algorithm
seems to be pretty powerful, even if it's memory hungry (for example,
cropmarks in Playback mode was trivial with this, and Canon buttons
are still visible). Also, it handles strange situations like when
Canon flash icon appears (it's almost in the middle of the screen).
Maybe I'll understand how ASM works someday :)
Alex
Did the text "Cropmks(1/3)" change to "CropmkP(1/3)" when pressing Q?
That P appears
when cropmarks in Playback mode are enabled.
K.
cristian paradiso
Oggetto: Re: [ML] [550D] False colors
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
cristian paradiso
Da: Alex <broscu...@gmail.com>
A: ml-d...@googlegroups.com
Inviato: Dom 23 gennaio 2011, 18:35:43
Oggetto: Re: [ML] [550D] False colors
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
cristian paradiso
Da: Alex <broscu...@gmail.com>
A: ml-d...@googlegroups.com
Inviato: Dom 23 gennaio 2011, 18:35:43
Oggetto: Re: [ML] [550D] False colors
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
Alex
>>> ml-devel+u...@googlegroups.com
>>> For more options, visit this group at
>>> http://groups.google.com/group/ml-devel?hl=en
>>
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+u...@googlegroups.com
K.
To unsubscribe from this group, send email to ml-devel+u...@googlegroups.com
cristian paradiso
Oggetto: Re: [ML] [550D] False colors
K.
cristian paradiso
cristian paradiso
Da: cristian paradiso <gui...@yahoo.it>
A: ml-d...@googlegroups.com
Inviato: Dom 23 gennaio 2011, 18:49:03
K.
cristian paradiso
JaKoB
JaKoB
Alex
for HDMI will also fix spotmeter.
On Sun, Jan 23, 2011 at 10:05 PM, JaKoB <jako...@gmail.com> wrote:
> but false color is reading fine thats how i find out the meter was not
> working right for hdmi
>
cristian paradiso
>>> For more options, visit this group at
>>> http://groups.google.com/group/ml-devel?hl=en
>>
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
M.v.M.
okay ? mvm
"cristian paradiso" <gui...@yahoo.it> schrieb:
cristian paradiso
cristian paradiso
Da: M.v.M. <Matthias_...@t-online.de>
A: ml-d...@googlegroups.com
Inviato: Dom 23 gennaio 2011, 20:19:00
Piers
Antony Newman
There was some debate about the these levels being clipped with I think was due to a quicktime bug (that someone on net updated an old old thread to say that Quicktime had been fixed).
The threads that I have read (and I have not read many), suggests that the Canon has a lot of data hidden outside the IRE range, and this data can be usefully recovered.
Based on this ... I have mapped the False colours to the entire Range.
I'll step back at the this point and let the debate continue.
AJ
Alex
Actually, I have a similar issue with 550D videos (clipped
brightness), but thought it was a problem with my codecs.
1: what I see in mplayer/vlc under Linux
2: what I get with vlc under Windows, normal still picture, silent
picture and frame export in avidemux
1 is much darker.
That's the reason I've set 10-245 limits on zebra by default...
K.
K.
M-L-Userrr
footage should look like and version 10 is wrong, there is less info
in dark areas,its black there and shouldnt be so black cause there is
some detail there.They might prefere version 10 but you can crush
black yourself if you want to,its not nice to be forced to do it by
bug in codec.
On 23 Sty, 21:13, "K." <justynadric...@gmail.com> wrote:
> http://www.dvinfo.net/forum/canon-eos-5d-mk-ii-hd/431290-whats-going-...
>
> 2011/1/23 K. <justynadric...@gmail.com>
>
>
>
>
> > This is bug in quicktime gamma and some h.264 implementations in players
> > /editors,i have it in players when not using quicktime 7.6 codec , in 7.5
> > version and older versions this bug exists ,its fixed in version 7.6 , it
> > exists again in quicktime x.Movies should look like on camera lcd.
> > Im not bothered that my player has this bug but i had to install proper
> > quicktime to get not crushed gamma in virtualdub .
>
> >> Welcome back!
>
> >> Actually, I have a similar issue with 550D videos (clipped
> >> brightness), but thought it was a problem with my codecs.
>
> >> 1: what I see in mplayer/vlc under Linux
> >> 2: what I get with vlc under Windows, normal still picture, silent
> >> picture and frame export in avidemux
>
> >> 1 is much darker.
>
> >> That's the reason I've set 10-245 limits on zebra by default...
>
> >> On Sun, Jan 23, 2011 at 10:52 PM, Piers <pie...@gmail.com> wrote:
> >> > A quick note regarding IRE
> >> > (Oh, and Hi, I'm back. Downloaded the focus assist build while I was
> >> away
> >> > and it's very cool).
> >> > I'm coming to the conclusion that the encoded movies are clipped to the
> >> > 16-235 Y range (which is the normal range for 8-bit digital video*) - aj
> >> (I
> >> > think) has said something like this, so hopefully if we're using his
> >> values
> >> > they already account for this. There's also some notes over at
> >> > I shot on holidays looks somewhat more overexposed than I would have
> >> > expected.
> >> > So (if I'm right), the formula from going from 0-255 Y to IRE would be:
> >> > IRE = (Y-16)/219
> >> > (IRE can be negative or >100, though these values should be clipped by
> >> the
> >> > display hardware)
> >> > PG
> >> > * the reason for this can be argued as "it's because it needed to run in
> >> > parallel with analogue video which can and does run outside the 0-100
> >> range"
> >> > or "because video engineers are dangerous idiots who never should have
> >> been
> >> > allowed near computer video"
>
> >> > --
> >> >http://magiclantern.wikia.com/
>
> >> > To post to this group, send email to ml-d...@googlegroups.com
> >> > To unsubscribe from this group, send email to
> >> >http://groups.google.com/group/ml-devel?hl=en
>
> >> --
> >>http://magiclantern.wikia.com/
>
> >> To post to this group, send email to ml-d...@googlegroups.com
> >> To unsubscribe from this group, send email to
Ba...@aol.com
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
To unsubscribe from this group, send email to ml-devel+u...@googlegroups.com
Ba...@aol.com
New build:
- False color area is smaller and faster; also a downsampled (low-res)
version is available (and used by default)
- Shortcut key: Flash button in movie mode, DOF preview in photo mode
(enabled by default, Q to disable)
Other changes:
- Fix from Deti: "This patch fixes focus peaking with HDMI display in
record mode."
- Cropmarks available in Playback mode (use Q to enable this)
- IRE spotmeter (from AJ's excel)
Also, Sawomedia created a page with many cropmarks:
http://www.sawomedia.de/magiclantern/cropmarks/
Vistor_Q
embarrassing, what's the use of the false color?
Thanks in advance...
On 23 ene, 15:42, cristian paradiso <guit...@yahoo.it> wrote:
> NO it didnt
>
> ________________________________
> Da: Alex <broscutama...@gmail.com>
> A: ml-d...@googlegroups.com
> Inviato: Dom 23 gennaio 2011, 18:35:43
> Oggetto: Re: [ML] [550D] False colors
>
> > Sorry Alex cropmarks in playback mode seems not working, I press Q but nothing
> >happens
>
> Did the text "Cropmks(1/3)" change to "CropmkP(1/3)" when pressing Q?
> That P appears
> when cropmarks in Playback mode are enabled.
>
>
>
>
>
>
>
>
>
>
>
> On Sun, Jan 23, 2011 at 8:33 PM, Alex <broscutama...@gmail.com> wrote:
> > Thanks, AJ. What I'd like is to integrate it with the mirror
> > algorithm, to avoid overwriting other stuff (here, the focus
> > rectangle, the spotmeter and part of histogram; maybe the Canon side
> > menus, if I'll expand the area).
>
> > That's the main reason for not using it by default. The algorithm
> > seems to be pretty powerful, even if it's memory hungry (for example,
> > cropmarks in Playback mode was trivial with this, and Canon buttons
> > are still visible). Also, it handles strange situations like when
> > Canon flash icon appears (it's almost in the middle of the screen).
>
> > Maybe I'll understand how ASM works someday :)
>
> > On Sun, Jan 23, 2011 at 8:07 PM, Antony Newman <antony.new...@gmail.com>
> >>=
> >> // Pix1a
>
> >>//======================================================================= =====================
> >>=
>
> >> " AND r14,r7,#0x0000FF00\n" // Luma of Vpixel 1 (shifted)
> >> " MOV r14,r14,LSR#6\n" // Luma x 4 (ie create word
> >> position to add)
>
> >> " LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
> >> " ORR r6,r6,r5,LSR#24\n" // add to bottom (1st) byte of
> >> Overlay colour
>
> >>//======================================================================= =====================
> >>=
> >> // Pix1b
>
> >>//======================================================================= =====================
> >>=
>
> >> " AND r14,r7,#0xFF000000\n" //
> >> " MOV r14,r14,LSR#24-2\n" // Luma x 4 (ie create word
> >> position to add)
>
> >> " LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
>
> >> " AND r5,r5,#0xFF<<24\n" // r5 = top byte = False colour ->
> >> get it
> >> " ORR r6,r6,r5,LSR#16\n" // add to 2nd byte of Overlay
> >> colour
>
> >>//======================================================================= =====================
> >>=
> >> // Pix2a
>
> >>//======================================================================= =====================
> >>=
>
> >> " AND r14,r8,#0x0000FF00\n" //
> >> " MOV r14,r14,LSR#6\n" // Luma x 4 (ie create word
> >> position to add)
>
> >> " LDR r5,[r0, r14]\n" // r5 = False Col + Hist value
>
> >> " AND r5,r5,#0xFF<<24\n" // r5 = top byte = False colour ->
> >> get it
> >> " ORR r6,r6,r5,LSR#8\n" // add to bottom (3rd) byte of
> >> Overlay colour
>
> >>//======================================================================= =====================
> >>=
> >> // Pix2b
>
> >>//======================================================================= =====================
> >> --
> >>http://magiclantern.wikia.com/
>
> >> To post to this group, send email to ml-d...@googlegroups.com
> >> To unsubscribe from this group, send email to
> >> ml-devel+u...@googlegroups.com
> >> For more options, visit this group at
> >>http://groups.google.com/group/ml-devel?hl=en
>
> --http://magiclantern.wikia.com/
Ba...@aol.com
Piers
- QuickTime Gamma - dark areas becoming darker
- Clipping - bright and dark info getting chopped off.
Alex
According to this topic, the luma -> IRE transformation is:
(16...235) -> (7.5...100)
http://cinema5d.com/viewtopic.php?f=24&t=346#p4673
Would it be OK to make an IRE spotmeter according to AJ formula and
another according to Piers? That's because I don't know who is right
:D
On Sun, Jan 23, 2011 at 10:52 PM, Piers <pie...@gmail.com> wrote:
Alex
This is equivalent to (0..255) -> (0.74, 108.44).
Alex
formula. I'd like to keep only one (which one?)
Also, I've got rid of "this function is not available in movie mode",
and applied latest patch from Deti regarding zebras on external
displays. Spotmeter is not yet fixed.
Alex
False colors still use AJ formula.
Fernando Freire
Alex
They have sharp teeth now :D
Does it happen on the external monitor only, or on built-in display
too? I did notice this a few days ago, but thought it was fixed...
On Mon, Jan 24, 2011 at 10:10 PM, cristian paradiso <gui...@yahoo.it> wrote:
> Hi Alex, this build it's great and It is wonderful to see zebras on my
> monitor and they are very fast. I have only one problem that wasnt there the
> previous firmware, The cropmarks are 1 or 2 pixel shifted to the right so
> you can see that on the left side there is a hole of 1/2 or 3 pixel. and one
> more thing, sometimes seems that the zebras make holes on the cropmarks :)
> strange I know, it makes holes that you can see trough. And I also dont
> understand how to remove the blu zebras leaving only the red ones. Thanks a
> lot for all your effort guys
> ________________________________
> Da: Alex <broscu...@gmail.com>
> A: ml-d...@googlegroups.com
> Inviato: Lun 24 gennaio 2011, 17:49:42
> Oggetto: Re: [ML] [550D] False colors
>
cristian paradiso
>>>> For more options, visit this group at
>>>> http://groups.google.com/group/ml-devel?hl=en
>>>
>>
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
Alex
>>>>> ml-devel+u...@googlegroups.com
>>>>> For more options, visit this group at
>>>>> http://groups.google.com/group/ml-devel?hl=en
>>>>
>>>
>>
>> --
>> http://magiclantern.wikia.com/
>>
>> To post to this group, send email to ml-d...@googlegroups.com
>> To unsubscribe from this group, send email to
>> ml-devel+u...@googlegroups.com
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>>
>> --
>> http://magiclantern.wikia.com/
>>
>> To post to this group, send email to ml-d...@googlegroups.com
>> To unsubscribe from this group, send email to
>> ml-devel+u...@googlegroups.com
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+u...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+u...@googlegroups.com
Piers
- 7.5 is the American NTSC black level (called "pedestal"). In Asian NTSC and everywhere else, 0 is black. See, e.g., http://en.wikipedia.org/wiki/NTSC-J
- I'm 99% sure that for ALL digital video types, 0 is black even in the americas. (Playback hardware is supposed to add the +7.5 to any analogue signals it makes)
- I strongly suspect that Canon would be using the full range for stills, and the clipped video range for movies, so we'll have to have both scales to be accurate
Cristian
Sent from my iPhone
Piers
lightimages
IRE, the AJ formula is in agreement with my software and hardware
scopes at every point on the chart. Pier's formula is not in
agreement at any point on the chart.
On Jan 24, 9:49 am, Alex <broscutama...@gmail.com> wrote:
> Here's a build which tests the difference between AJ and Piers' IRE
> formula. I'd like to keep only one (which one?)
>
> Also, I've got rid of "this function is not available in movie mode",
> and applied latest patch from Deti regarding zebras on external
> displays. Spotmeter is not yet fixed.
>
>
>
>
>
>
>
> On Mon, Jan 24, 2011 at 7:39 PM, Alex <broscutama...@gmail.com> wrote:
> >> (16...235) -> (7.5...100)
> > This is equivalent to (0..255) -> (0.74, 108.44).
>
> >> Piers,
>
> >> According to this topic, the luma -> IRE transformation is:
> >> (16...235) -> (7.5...100)
> >>http://cinema5d.com/viewtopic.php?f=24&t=346#p4673
>
> >> Would it be OK to make an IRE spotmeter according to AJ formula and
> >> another according to Piers? That's because I don't know who is right
> >> :D
>
> >> On Sun, Jan 23, 2011 at 10:52 PM, Piers <pie...@gmail.com> wrote:
> >>> A quick note regarding IRE
> >>> (Oh, and Hi, I'm back. Downloaded the focus assist build while I was away
> >>> and it's very cool).
> >>> I'm coming to the conclusion that the encoded movies are clipped to the
> >>> 16-235 Y range (which is the normal range for 8-bit digital video*) - aj (I
> >>> think) has said something like this, so hopefully if we're using his values
> >>> they already account for this. There's also some notes over at
> >>> expected.
> >>> So (if I'm right), the formula from going from 0-255 Y to IRE would be:
> >>> IRE = (Y-16)/219
> >>> (IRE can be negative or >100, though these values should be clipped by the
> >>> display hardware)
> >>> PG
> >>> * the reason for this can be argued as "it's because it needed to run in
> >>> parallel with analogue video which can and does run outside the 0-100 range"
> >>> or "because video engineers are dangerous idiots who never should have been
> >>> allowed near computer video"
>
> >>> --
> >>>http://magiclantern.wikia.com/
>
> >>> To post to this group, send email to ml-d...@googlegroups.com
> >>> To unsubscribe from this group, send email to
> >>> ml-devel+u...@googlegroups.com
> >>> For more options, visit this group at
> >>>http://groups.google.com/group/ml-devel?hl=en
>
>
>
> 1146KViewDownload
Antony Newman
Alex: It only takes 1KB per table, just store both and let the users decide.
Lightimages: OMG .. where would I be now if you wrote my school reports!
AJ
K.
Hey Alex,
there is a screenshot of the 2/3 pixel shifted cropmarks
Inviato: Lun 24 gennaio 2011, 20:58:57
>>>>> ml-devel+unsub...@googlegroups.com
>>>>> For more options, visit this group at
>>>>> http://groups.google.com/group/ml-devel?hl=en
>>>>
>>>
>>
>> --
>> http://magiclantern.wikia.com/
>>
>> To post to this group, send email to ml-d...@googlegroups.com
>> To unsubscribe from this group, send email to
>> ml-devel+unsub...@googlegroups.com
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>>
>> --
>> http://magiclantern.wikia.com/
>>
>> To post to this group, send email to ml-d...@googlegroups.com
>> To unsubscribe from this group, send email to
>> ml-devel+unsub...@googlegroups.com
>> For more options, visit this group at
>> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+unsub...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
>
> --
> http://magiclantern.wikia.com/
>
> To post to this group, send email to ml-d...@googlegroups.com
> To unsubscribe from this group, send email to
> ml-devel+unsub...@googlegroups.com
> For more options, visit this group at
> http://groups.google.com/group/ml-devel?hl=en
--
http://magiclantern.wikia.com/
To post to this group, send email to ml-d...@googlegroups.com
To unsubscribe from this group, send email to ml-devel+unsub...@googlegroups.com