I've read a few comments about Toshiba cell processor "super upconversion" of 480i material using 9 frames to create a 960p image. While I'm guessing the hype is greater than reality, I'm wondering how their techniques compare to those used in the current generation of video processors? Personally, I am viewing a smaller percentage of SD material each year although I would be interested in a technique which provided a better picture. I'm somewhat skeptical as both my VP50 and PS3 at 1080p both look pretty good.

Well, the PlayStation 3 is using the Power PC "cell processor" as its video processing engine. Didn't realize that Toshiba was using same.

What Toshiba product are you talking about exactly?

PS3 Upconversion is really good.

I've seen the units above. Check out the Samsung 5000. I took it back but the upcoversion was so good I am going to get another one as soon as a few more of the unit's bugs and firmware updates are addressed. I had the 5K for a couple of weeks and really enjoyed looking at old sd DVDs. I felt it was by far the best upconversion I have seen. It uses the Reon chip and I understand the Toshiba XA2 also does but I really felt the 5K was better.

I've read a few comments about Toshiba cell processor "super upconversion" of 480i material using 9 frames to create a 960p image. ... I'm wondering how their techniques compare to those used in the current generation of video processors? Personally, I am viewing a smaller percentage of SD material each year although I would be interested in a technique which provided a better picture. I'm somewhat skeptical as both my VP50 and PS3 at 1080p both look pretty good.
I don't think any visible improvement is had using more than 4 consecutive fields to do de-interlacing with. Actually the material to construct a de-interlaced frame is taken from at most three fields (current, previous, and next). A fourth and possibly a fifth might be used to help the processor decide which of the aforementioned three to take material from.

After the 480i is converted to 480p, the de-interlaced 480p frames would be upscaled one at a time if 960p or other higher resolution capable format is desired.

Video hints: http://members.aol.com/ajaynejr/viddoubl.htm

Allan,

I don't believe it works like that. I think the idea is to take information about the same object from multiple frames and use the data from all those frames to compose the final image. Each frame can have fractional-pixel phase shifts in position of the objects, and that data can be used to extract additional resolution for the final rendered image. In other words, it's not just a matter of deciding which field out of two or three to take a pixel from, but rather analyzing a large number of fields and using the total information to construct a higher-resolution final image.

- Dale Adams

IVTC uses 3 fields generally, but deinterlacing static content can use as many as you want.