<div dir="ltr">Comparing the NV12 image with the RGB I can readily see it is the color-space conversion that causes the distortion--not libva or the intel driver. My apologies for the diversion.</div><div class="gmail_extra"><br><div class="gmail_quote">On Wed, Oct 8, 2014 at 7:35 PM, Zhao, Yakui <span dir="ltr"><<a href="mailto:yakui.zhao@intel.com" target="_blank">yakui.zhao@intel.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class="">On Wed, 2014-10-08 at 10:02 -0600, Matt Pekar wrote:<br>
> I've examined the behavior on Intel(R) Core(TM) i3-4130 and Intel(R)<br>
> Core(TM) i5-2400. We target resolutions as low as 224x64 and up to<br>
> 720P.<br>
><br>
> Another thing I'm noticing is that the colors used seem to matter. I<br>
> put some white on black text up and it looks great. Maybe there is an<br>
> issue with how I do my colorspace conversion from BGR to NV12...<br>
><br>
<br>
</span>Hi, Matt<br>
<br>
In order to narrow down the issue, Will you please help to remove the<br>
extra color-space conversion of NV12 -> RGB and instead use the NV12<br>
image directly for the encoding?<br>
<br>
After the encoding is finished, you can calculate the corresponding PSNR<br>
with the original reference YUV and then see whether the PSNR can meet<br>
with your requirement.<br>
<br>
Thanks.<br>
<span class="HOEnZb"><font color="#888888"> Yakui<br>
</font></span><div class="HOEnZb"><div class="h5"><br>
> I'm going to attempt to attach two pictures of what I'm seeing. They<br>
> show the original .png image and what the encoder delivers side by<br>
> side.<br>
><br>
><br>
> On Wed, Oct 8, 2014 at 10:41 AM, Chris Healy <<a href="mailto:cphealy@gmail.com">cphealy@gmail.com</a>><br>
> wrote:<br>
> Hi Matt,<br>
><br>
><br>
> I've been using the libva with the Intel driver to encode and<br>
> stream myself for quite some time now. There are a number of<br>
> visual artifacts that I have encountered along the way, though<br>
> the fuzziness you describe is not one of them. With the Intel<br>
> driver, I would expect different visual behaviour depending on<br>
> the particular HW you are using as each one is different. I'm<br>
> using an i7-3517U to encode 1280x720 H.264 at 4Mbps. What are<br>
> you working with?<br>
><br>
> Chris<br>
><br>
><br>
> On Wed, Oct 8, 2014 at 7:18 AM, Matt Pekar<br>
> <<a href="mailto:mpekar@raineyelectronics.com">mpekar@raineyelectronics.com</a>> wrote:<br>
><br>
> We use libva with the Intel driver to do live video<br>
> streaming. In the encoding process we dynamically<br>
> ramp the QP up and down as different content flows<br>
> through.<br>
><br>
><br>
> There are cases where we display simple text messages<br>
> for many seconds. In these cases I'd like to send my<br>
> QP down to 1--highest quality--and get as close to a<br>
> lossless encode as possible.<br>
><br>
><br>
> What I'm seeing is that on fonts and sharp lines a<br>
> little bit of encoding will be done regardless,<br>
> leaving the text slightly fuzzy. I tried the<br>
> new VAEncMiscParameterTypeQualityLevel setting, but it<br>
> only had two levels (1 and 2) and they didn't seem to<br>
> affect the picture at all.<br>
><br>
><br>
> Are there any settings I might try to ramp the quality<br>
> up even further? Is the fuzziness I'm seeing in the<br>
> text inherent to the h264 standard, or is this just<br>
> what Intel's implementation happens to do?<br>
><br>
><br>
><br>
><br>
><br>
> _______________________________________________<br>
> Libva mailing list<br>
> <a href="mailto:Libva@lists.freedesktop.org">Libva@lists.freedesktop.org</a><br>
> <a href="http://lists.freedesktop.org/mailman/listinfo/libva" target="_blank">http://lists.freedesktop.org/mailman/listinfo/libva</a><br>
><br>
><br>
><br>
><br>
><br>
<br>
<br>
</div></div></blockquote></div><br></div>