hi,
I'm using DM6467T to implement live video encoding with TVP5147 composite input as PAL(720x576 25fps) video capture.
After investigating the DMAI source and testing, i found that the picture(YUV 4:2:2 semi-planar) captured by ioctl is a complete video frame:720x576 Y pixels, 720x576/2 pixels in each U and V. In addition, there exists obvios interline twitter or combing effect in the motion part of each picture, but still part of the picture is good.
I guess the tvp5147 driver should have handled the deinterlacing of 2 fields when capture interface called. But frankly the deinterlace result is not satisfied. The encoding to these deficient frames produce the video that motion parts have interline twittering and long-tail scene.
Anybody can give me some advice how to handle the problem? Very eager to get it work and your reply is highly expected and appreciated.
BTW, we ported the X264 into DSP with a lot of optimization including time-comsuming module linear assembly. This causes the DSP works very hard and the CPU load always approximates 100%. Hope the deinterlacing to the YUV422semi or YUV420semi converted by VDCE is simple and light-weight, if runnng in ARM, it would be best.