On date Wednesday 2010-06-02 14:34:15 +0200, Michael Niedermayer wrote:
On Mon, May 31, 2010 at 06:17:00PM +0200, Stefano Sabatini wrote:
On date Friday 2010-05-28 03:23:38 +0200, Michael Niedermayer wrote:
On Thu, May 27, 2010 at 12:14:24PM +0200, Stefano Sabatini wrote:
On date Thursday 2010-05-27 00:59:21 +0200, Stefano Sabatini wrote: [...]
I was almost committing, but since there are other formats I want to add support for, maybe someone can find out a scheme for all of them.
These are the format I want to add support for: bgr4_byte ->none rgb4_byte ->none rgb444be ->none rgb444le ->none bgr444be ->none bgr444le ->none
I propose this scheme.
First byte: b = BGR B = BGR+A r = RGB R = RGB+A a = ABGR A = ARGB
i dont like this at all, using upper/lower case to distinguish formats
following is much closer to the existing system: RGBA -> RBA[bits] ARGB -> ARB[bits] ABGR -> ABR[bits] BGRA -> BRA[bits]
This only works if all the components have the same amount of bits.
Anyway I don't mind if you want to use this scheme for the missing formats: rgb444be -> [16]BRA rgb444le -> ARB[16] bgr444be -> [16]RBA bgr444le -> ABR[16]
the bits value should be 64 for 16bit per component, 32 for 8bpc
This would lead to: rgb444be -> [16]BGR rgb444le -> RGB[16] bgr444be -> [16]BGR bgr444le -> BGR[16] note that rgb444/bgr444 does not use the A component, which is set to 0 or to 1 (check libavutil/pixfmt.h): PIX_FMT_RGB444BE, ///< packed RGB 4:4:4, 16bpp, (msb)4A 4R 4G 4B(lsb), big-endian, most significant bits to 0 PIX_FMT_RGB444LE, ///< packed RGB 4:4:4, 16bpp, (msb)4A 4R 4G 4B(lsb), little-endian, most significant bits to 0 PIX_FMT_BGR444BE, ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), big-endian, most significant bits to 1 PIX_FMT_BGR444LE, ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), little-endian, most significant bits to 1 but that shouldn't be a problem, since I can't see a reason for which it shouldn't just be ignored. Of the four listed above two are already used: RGB[16] Packed RGB 5:6:5, 16bpp, (msb) 5R 6G 5B(lsb), little-endian [NOT in AVI] BGR[16] Packed BGR 5:6:5, 16bpp, (msb) 5B 6G 5R(lsb), little-endian [NOT in AVI] That's why I proposed the other scheme, which would allow to define without ambiguity also formats for which the number of bits per component is different (but only works if bits_per_component <= 9). Alternatively we could use lowcased characters: rgb -> RGB bgr -> BGR rba -> RGBA arb -> ARGB abr -> ABGR Check solution in attachment, regards.