Bugzilla – Bug 1203
RGB webcams
Last modified: 2004-11-18 02:27:03
You need to log in before you can comment on or make changes to this bug.
My webcam (Kensington 67014) which uses the SE401 driver only produces RGB24 frames. I've patched vic to make it check if YUV production is supported, and if not it falls back to RGB24. The patch to video/grabber-video4linux.cpp follows: Index: grabber-video4linux.cpp =================================================================== RCS file: /cvs/fl/ag-media/vic/video/grabber-video4linux.cpp,v retrieving revision 1.2 diff -r1.2 grabber-video4linux.cpp 47a48,49 > #include "rgb-converter.h" > 71a74 > #define CF_RGB24 3 313a317 > bool force_rgb24 = false; 352a357,361 > vid_mmap.format = VIDEO_PALETTE_YUV422; > if (-1 == ioctl(fd_, VIDIOCMCAPTURE, &vid_mmap)){ > force_rgb24 = true; > } > 408,409d416 < < 417c424,426 < --- > if(force_rgb24) > cformat_ = CF_RGB24; > 569,571c578,580 < if (-1 == ioctl(fd_, VIDIOCMCAPTURE, &gb_even)) < perror("ioctl VIDIOCMCAPTURE even"); < else --- > if (-1 == ioctl(fd_, VIDIOCMCAPTURE, &gb_even)){ > perror("V4lGrabber::start() ioctl VIDIOCMCAPTURE even"); > } else 574,576c583,585 < if (-1 == ioctl(fd_, VIDIOCMCAPTURE, &gb_odd)) < perror("ioctl VIDIOCMCAPTURE odd"); < else --- > if (-1 == ioctl(fd_, VIDIOCMCAPTURE, &gb_odd)){ > perror("V4lGrabber::start() ioctl VIDIOCMCAPTURE odd"); > }else 609c618 < perror("ioctl VIDIOCSYNC"); --- > perror("V4lGrabber::grab() ioctl VIDIOCSYNC"); 632a642,647 > > case CF_RGB24: > RGB_Converter *converter = new RGB_Converter_411(24,NULL,0); > converter->convert((u_int8_t*)fr, width_, height_,(u_int8_t*)frame_,width_,height_,0); > delete converter; > break; 634c649 < --- > 637,639c652,654 < (grab_count % 2) ? &gb_odd : &gb_even)) < perror("ioctl VIDIOMCAPTURE"); < else --- > (grab_count % 2) ? &gb_odd : &gb_even)){ > perror("V4lGrabber::grab() ioctl VIDIOMCAPTURE"); > }else 816c831,835 < v4lformat_=VIDEO_PALETTE_YUV422; --- > if (cformat_ == CF_RGB24) { > v4lformat_= VIDEO_PALETTE_RGB24; > } else { > v4lformat_ = VIDEO_PALETTE_YUV422; > } 840a860,863 > case CF_RGB24: > set_size_411(width_, height_); > DEBUG(fprintf(stderr," RGB24")); > break; 841a865 >
Created an attachment (id=86) [details] Falls back to RGB if YUV fails
The patch breaks webcams based on the pwc driver (or more generally devices that only support YUV420). I think the following code in the patch: vid_mmap.format = VIDEO_PALETTE_YUV422; if (-1 == ioctl(fd_, VIDIOCMCAPTURE, &vid_mmap)){ force_rgb24 = true; } should be: vid_mmap.format = VIDEO_PALETTE_RGB24; if (-1 != ioctl(fd_, VIDIOCMCAPTURE, &vid_mmap)){ force_rgb24 = true; } Which makes the pwc based webcams work once again. I would also recommend that the converter local variable in the following code be a member variable of the V4lGrabber class, to avoid C++ overheads of calling the RGB_Converter_411 C++ constructor and destructor up to 30 times a second: int V4lGrabber::grab() { ... case CF_RGB24: RGB_Converter *converter = new RGB_Converter_411(24,NULL,0); converter->convert((u_int8_t*)fr, width_, height_,(u_int8_t*)frame_,width_,height_,0); delete converter; break; } ... } Doug
If this code is used: vid_mmap.format = VIDEO_PALETTE_RGB24; if (-1 != ioctl(fd_, VIDIOCMCAPTURE, &vid_mmap)){ force_rgb24 = true; } It would mean that if the camera FAILS to work in RGB24 mode, then force it into RGB24 mode. The reason I checked against VIDEO_PALETTE_YUV422 is so that if it failed then it would fall back. Unfortunately I don't have multiple cameras so I can't really test this patch as much as I would like.
I think the logic is the other way around, as I changed it from == -1 to != -1 in the ioctl test, so it should only be set to RGB24 if the ioctl test for RGB24 doesn't fail. Doug
Created an attachment (id=87) [details] new RGB24 patch The attached patch is a combination of both, it will only force RGB24 if supported and then only if YUV422 isn't. It should hopefully be a suitable solution.
