Execution ended after 0:00:06.814207550 Setting pipeline to NULL. New clock: GstSystemClock ^Chandling interrupt.
Pipeline is live and does not need PREROLL. Enabling this feature on C2 produces similar results (eliminate key frame "popping").Ģ) QP / quant matrix - using the same algorithm for rate control/quality would likely produce similar results.Gst-launch-1.0 v4l2src device=/dev/video0 ! image/jpeg,width=1920,height=1080,framerate=3000/100 ! jpegdec ! avenc_h264_omx ! fakesink Setting pipeline to PAUSED. In conclusion, the issues identified are:ġ) GOP - ffmpeg is likely using "automatic" GOP detection. Integrating the Amlogic encoder into ffmpeg would likely achieve similar results: increased quality + longer encode times. This would account for the time discrepancy between the two.
The RPi is likely using ffmpeg (non-realtime) to determine QP and quant matrix. My theory is that C2 uses it own rate control code (realtime) to determine QP. This narrows the scope of the issue down to QP (quantization parameter). The software decoding also guarantees identical input. Based on file size, the bitrate is comparable.
The bitstream indicates that only I and P frames are used. That answers two important questions: Are B frames used? Is the bitrate comparable? Fretzke wrote:…and the resulting file from RPi with software decoding:
0 kommentar(er)
