[StinDaWg]

For anyone who thinks a sharpened image is better this is a good read http://www.videophile.info/Guide_EE/Page_01.htm

first: i dont talk about edge enhancement. there are a lot of sharpen methods, i talk about "unsharp masking", thats completly different.

second: with unsharp masking i can confirm 100% what StinDaWg wrote: "Any softness in the picture was cleared up. Facial details that I couldn't see before like pores, freckles, hair, all became suddenly crisp like I was looking at native 1080p content."


as i sayd: i had a kathrein tv-receiver whitch is known for good picture quality. as i build my htpc i used at first xbmc 12 and compared the picture of livetv between receiver and xbmc by one klick on my remote. the different was huge. the receiver had a much sharper picture and more powerfull color. after that i tried mp+lav+dvxa2 nativ without postprocessing. picture was compared to xbmc a little sharper than xbmc but as not as sharp as tv receiver. colors was much better than xbmc, like the tv receiver. after that i used mp+lav+ffdshow (sharpen, unsharp masking) and that was the only way i get a picture that is 99% as good as the picture of the kathrein receiver. you have to set the amount of sharpnes carefull, not to much and in depency of the video input resolution.

Can you use MPC-HC shaders with DSPlayer, or will only ffdshow work? I don't have a powerful pc, and I know ffdshow uses cpu while MPC-HC uses GPU for post-processing. The default sharpen complex works good without having to mess around with 100 different settings.

[Ed76]

this should be possible: http://forum.xbmc.org/showthread.php?tid=101097

yes, ffdshow sometimes is a little bit complicated but if you want to try ffdshow i can tell you that the sharpen methods dont need much cpu power.

[StinDaWg]

It's not possible to use post-processing with DXVA on ffdshow is it? I tried it out but every time I inserted the re-sizing/sharpen filters into MPC-HC, it disabled DXVA and cpu shot up to 60-80% just on a 720p encode.

[Ed76]

for ffdshow postprocessing you have to use dxva2 copy back. copy back needs a little bit more cpu power than native but less power than software mode.

[StinDaWg]

for ffdshow postprocessing you have to use dxva2 copy back. copy back needs a little bit more cpu power than native but less power than software mode.

Thanks, it's still using 40-60% cpu on DXVA copyback, but it's running smooth so that's all that matters. I've been trying out ffdshow resize and I think I like it better than the unsharp masking shaders. Using those shaders changes the contrast and color of the picture a little, which I don't really like. ffdshow method seems more like a straight forward resize and sharpen without messing with the picture otherwise. I also noticed on the unsharp mask shaders it causes the top and bottom pixels of the video to be kind of blurred or glossy looking. This is really noticeable when you are playing a movie with black bars at the top and bottom.

Right now I'm using ffdshow resize to 1920x, Bicubic and 1.5 luma sharpen, which looks pretty good. What are you using?

[oldpoem]

Right now I'm using ffdshow resize to 1920x, Bicubic and 1.5 luma sharpen, which looks pretty good. What are you using?

I'm using resize to desktop resolution, spline chroma, spline luma , swscaler 0.6 sharpen

[StinDaWg]

Played around with ffdshow all day, and read some forum posts around the net. I settled on Lanczos4 and 1.0 luma sharpen. Looks really good. Now off to read the DSPlayer thread and see if it's worth it to try to figure that beast out.