HD Camera Resolution

chrisM · June 27, 2004

I have a few questions regarding the VERTICAL resolution of the Sony and Panasonic HD cameras. I am not interested in theoretical resolution, but "real world" resolution gained from shooting a resolution chart and examining it, as you would for film, (perhaps using a chart designed for digital still cameras like the 13A/ISO 12233)

My questions are:

1. What vertical resolution can be seen, or measured, in 1080i, 1080p and/or 720p

2. Is there a significant difference between 1080i and 1080 24P on Sony HD with a static test subject?

3. How does vertical res compare to horizontal res, relative to picture height?

Also, does anyone know if the Sony 1080i cameras have an optical anti-aliasing filter, or only a "software" filter, and do these cameras combine pixels, (in alternate lines on alternate fields), like a Digi Beta, when in interlace mode?

Thanks for your assistance.

Jason Rodriguez · June 27, 2004

I believe according to a test from *Kodak* (in other words, not Sony; this is sort like Apple running a speed comparison test against a Dell) that the Cinealta had a verticle resolution around 860 lines-per-picture-height (using the ISO 12233 chart). The film (scanned at 4K) was around 1700 lines-per-picture-height. Of course I must commend Kodak because this is the only time I've ever seen the two mediums compared using the same ISO chart and not company specs, so FWIW.

David Mullen ASC · June 27, 2004

All the Sony HD cameras have an optical anti-aliasing filter as far as I know, but all I know for sure is that the progressive-scan ones do. You also have to consider whether you are talking about resolution before or after recording to HDCAM, which reduces 1920 to 1440 horizontally. Most tests, like Kodak's, have been based on what the camera can record to its own HDCAM deck, i.e. 1440 x 1080, 3:1:1, and fairly compressed overall.

Phil Rhodes · June 27, 2004

Hi,

Once you consider the tape format, the absolute resolution becomes impossible to define because of the way the compression works. A small area of fine detail in the middle of a frame of very low-frequency detail will be rendered more sharply than the same area surrounded by other areas of equally high-frequency detail. For example, if you frame up a resolution chart with small areas of lines surrounded mainly by white or solid blocks of colour/greyscale test areas, it will appear to resolve more lines than if you filled the entire frame with lines at the same pitch.

This occurs because the compression codec has to fit each frame into a given amount of data space on the tape, and therefore has to truncate the DCT coefficients of high-detail images more than low-detail ones. Compare this to many types of computerised video which have a dynamic frame size and are free to use more data space for more difficult frames.

The result is that the apparent sharpness of something like an F900 recording to its onboard HDCAM deck is entirely dependent on the content of the frame, and is impossible to absolutely quantify.

Phil

Luke Prendergast · June 27, 2004

had a verticle resolution around 860 lines-per-picture-height

Lines-per-picture-height is a measure of horizontal rather than vertical resolution isn't it?

chrisM · June 28, 2004

I believe according to a test from *Kodak* (in other words, not Sony; this is sort like Apple running a speed comparison test against a Dell) that the Cinealta had a vertical resolution around 860 lines-per-picture-height (using the ISO 12233 chart)

I assume this was with the camera operating in 1080P mode? Does this mean that resolution would be lower in 1080i? Tests I've done with a PAL DigiBeta (recorded to tape) show that about 430ish lines are achieved in field mode while 500+ can be resolved in frame (AVS) mode. In field mode the camera "reads out" all CCD pixels every field and alternately combines them, (i.e. field 1: 1+2, 3+4... field 2: 2+3, 4+5 etc) this has the effect of reducing resolution to 75% of the the theoretical maximum.

One way of seeing if there is a change in resolution is to switch between 1080i and 1080P output as an interlace frame, on a highly detailed static subject, and see if line flicker increases when displayed on an interlace display device.

Jason Rodriguez · June 28, 2004

I assume this was with the camera operating in 1080P mode? Does this mean that resolution would be lower in 1080i?

Theoretically the resolution should be the same, although only when moving, and when viewed on an interlaced monitor. When viewed on a computer screen, if there's movement, then you've reduced the verticle resolution by half, because you must see a whole frame on the computer at any given time, and so in order to see no interlacing artifacts, you're going to have to throw away, or interpolate a full-frame out of an interlaced one.

But again, on a still shot, the resolution of both will be the same.

chrisM · June 28, 2004

But again, on a still shot, the resolution of both will be the same.

Have you measured this? 860 lines PPH should be enough to cause interline flicker issues, and I haven't noticed excessive interline flicker on 1080i material. Could it be that they reduce the maximum possible res in 1080P so it doesn't flicker excessively when displayed 1080Pif?

June 28, 2004

If the picture is static, interlaced and progressive resolution is the same.

chrisM · June 28, 2004

A Sony tech told me today that the vertical resolution in 1080i is less than that achieved with a static subject in 1080P.

The tech agreed that in interlace mode the HD cameras did vertical pixel "addition", i.e low pass filtering as I described earlier, to reduce interline flicker.

I don't know who is right, the Sony tech wasn't 100% sure. Has anyone done independent tests?

Phil Rhodes · June 28, 2004

Hi,

Many SD cameras (mine included) have a "super V" option which maximises vertical detail by turning off interline antialiasing. It flickers more, especially on motion, but the results are noticeably sharper.

Phil

June 28, 2004

I read somewhere that 1080i and 720p have the same perceived resolution. I thought that it was on moving images. Maybe it is some kind of average, or maybe it is even on staic images and on moving images it gets worse.

HDCAM has only about 1000 lines horizontal resolution.

Can someone explain to me all of this. I am a little lost.

Pete

chrisM · June 29, 2004

This article explains the issues regards 720P vs 1080i: progressive vs interlace

What is not analyzed in this document is whether 1080 25P (recorded as an interlaced frame) is better than 1080i. Some of the debate about 1080i is based on the resolution lost through interline flicker, i.e. even when the resolution is there in the recording you won't see it because the line flicker obscures it. However, the resolution probably isn't there to begin with if Sony adopt the same process with the HD Cams in interlace as they do for standard def cameras. This article Camera pdf(pg 2) gives a pretty good explanation of the process.

Mike Brennan · June 29, 2004

My questions are:

1. What vertical resolution can be seen, or measured, in 1080i, 1080p and/or 720p

2. Is there a significant difference between 1080i and 1080 24P on Sony HD with a static test subject?

3. How does vertical res compare to horizontal res, relative to picture height?

Do the test yourself it is easy!

One HD camera one resolution test chart two lights. Display on either HD monitor or projector. Look at the res lines on the chart.

If you have a possible job someone will transfer your test chart from HD to film.

A subject other than test chart should also be considered as comression schemes of HD recording affect resolution in unexpected ways, ie subject dependent.

What decisions will you be making based on viewing a test chart?

Mike Brennan

chrisM · June 29, 2004

I don't have access to an HD camera at the moment, I was just interested in people's opinions. I will have access soon and will do my own tests.

I shoot mainly on a DigiBeta 790 at present, but a lot of production in Australia is changing over to HD. 1080i is mandated for free to air broadcasts. I'm interested in "real world" resolution to determine display scaling losses etc., i.e. how much res you lose on a projection system that isn't 1080 native. My tests on a PAL Digi Beta really surprised me at how soft the vertical res really is when displayed on a progressive display device.

John Pytlak RIP · June 29, 2004

Data and pictures are published in the technical paper "Assessing the Quality of Motion Picture Systems from Scene-to-Digital Data" by Roger Morton, Michelle Maurer, and Chris DuMont, SMPTE Journal, February/March 2002, Volume 111, pages 85-96. There is a very significant difference in the ability to capture fine image detail, not to mention the much greater latitude and dynamic range of film.