SVCD video is pale

[XEPACKOB]

Hello,

When I am converting my videos (any -- captured from DV camcorder, from DVD, etc.) to SVCD resulting video is of a very bad quality: image is pale, instead of black I see a color closer to dark-grey. VCD conversion is just fine. I tried different videos , both tmpgenc and cinemacraft -- all the same.

Any help appretiated!!

[adam]

Short answer: When encoding DV in TMPGEnc go to the quantize matrix tab and check, "Output YUV data as Basic YCbCr not CCIR601."

Long answer: Camera's and pc's use colors which extend beyond the ranges which a tv can display. Out of a full range of 0-255 levels of luminence, a tv can only use about 16-235 (or 8-235 depending on region.) So when you create something on your pc, ie: animation, or you film something on a conventional camera, you have colors in the 0-255 range and if you were to then encode and display that on a tv, anything extending beyond the 16-235 range would be hard clipped. This results in very noticable artifacts. So the solution is to compress your luminence levels down to CCIR601 specifications, which is what the 16-235 range is. By default, TMPGenc does this.

Now, DV is actually already in the 16-235 range, so if you simply load this source into TMPGEnc and encode, then you further compress the luminence scale even though it already complies with CCIR601. You are simply throwing out the dynamic range of your colors and the most apparant effect is that you no longer get "true black" and instead get something closer to grey. Bright whites will also be washed out.

So generally speaking, you should enable the above mentioned option and your luminence ranges should be preserved, because they should already be correct as they are.

Even though I'm sure this is what's happening in your case, its important to realize that even though DV is filmed in the 16-235 range, many DV codecs will stretch it out to 0-255 without you even knowing it. The Canopus DV Codec does this for example, and you must compress the ranges to CCIR601 spec or you will get artifacts and in many cases VERY dark picture.

I think luminence levels are something that are far too commonly overlooked. By and large, unless you are dealing with non-DV camera footage or your own pc created images, ie: animation, then you are going to be dealing with footage that already conforms to CCIR601. The default setting in TMPGenc will further compress this range and you will be just throwing quality out the window. I think the majority of TMPGEnc users out there are probably using this setting incorrectly.

When dealing with conversions its imperative to know the color range of your source, what effect all of your software will have on it, and what color range your target device uses. Ideally, you want to preserve the luminence range as is throughout the entire process and then make any necessary compression during encoding.

[pbhalerao]

I noticed the same problem. What a precise answer!

We newbie's would have never known. Brilliant.