It seems that the image size was changed to 19201088(for 16-byte alignment) after processing by decoder. I thought nvvidconv should cropped the raw image to 19201080 according to the mata-info before processing, but in fact it just resized the image to 1920*1080 and this would cause the output image contains several color stripes .
I also made a test by using nvivafilter to instead of the nvvidconv, the output image is OK. but It seems that the nvivafilter takes high CPU performance, I don’t know why! The pipeline I used as follows:
I know that I can assign the cropped-rect in nvvidconv when processing the 1080p videos, But In fact I don't even know the video resolution that i used. In addition, considering that the video resolution may be changed, i can't set the cropped rectangle to a fixed value.
Maybe you don’t understand what i mean, i just want to remove the color stripes contained in decoded image.
I can’t use nvvidconv plugin to crop the image because i don’t know the video resolution. Any other ways to resolve this issue?
Thanks for your reply. As i said above, after i added nvviafilter while using omxh264dec,the CPU usage will become higher. In fact, I used the decodebin to auto-magically construct a decoding pipeline as i don’t know the video format(maybe a H265 format).
I utilized the ‘appsrc’ and push back the H264/H265 video frames to the gst pipeline to instead the ‘rtspsrc’.
Then i utilized ‘parsebin | nvv4l2decoder’ to instead the ‘decodebin’, it works well, thank you!