555 HD 1080i frame copy has ghosting

[hikent]

Hi,

I'm having a problem where there appears to be ghosting when I copy a frame (CTRL-C) & save as a BMP. I was trying to clean up a menu background that looked fuzzy. When I opened it in photoshop there was terrible ghosting of the image & even the HDNet bug. I opened it up in another free BMP editor & saw the same -- so I suspect something to do with the BMP file.

I'd like to upload the original BMP -- but I'm limited to 640 x 280 or something. The HD image is 1920x1080.

Any thought? I'm curious if it has something to do with de-interlacing.

Thanks. Kent

[Dan203]

Probably a interlacing problem. Try running it through the PhotoShop deinterlace filter and see if that helps at all. If not you should try stepping forward or back a couple of frames. Some frames show more interlacing artifacts then others.

Dan

[hikent]

Thanks for the advice -- I'll try that.

In general, are the frames with a P, I, or B better to grab? (And what do the letters stand for? I can't find this in the manual.)

Thanks. Kent

[bkh]

> (And what do the letters stand for? I can't find this in the manual.)

Here is the general idea but some details probably are wrong because at this point I'm recalling vague old memories:

An "I" frame is an "independent" frame. It encodes a complete picture that stands alone.

A "P" frame is a "predictive" frame. From the nearest 2 previous frames (I and/or P) they do motion prediction in the forward direction, and the next P frame just encodes the differences between what was predicted and what actually is in that frame of video. Given that the predictions are usually good, encoding the differences takes less space than encoding the full data of the frame (as an "I" frame).

A "B" frame is a "between" frame. They do forward and backward motion prediction between the nearest preceding and following frames (of types P and/or I? ...can't recall) and encode only the difference between the prediction and the actual frame contents. Since they predicted knowing both the past and future frames, the prediction should be better than just a forward prediction, so the B frame encoding is even smaller than a P.

[hikent]

Thanks for the replies. Between the deinterlace filter & selecting an I frame, the background is much improved. Thanks.