Originally Posted by stanger89
I'm actually beginning to think that the really important thing that's been/being done here, is not the gathering of new data for projectors, but the investigation of how the current data (On/Off, ANSI) relates to real-world images and projectors.
This intra-scene contrast calculator is an intriguing step forward and I thank Erik for putting it together. Because it only uses max and min luma to calculate intra-scene contrast though it seems like it could yield some unexpected results particularly with a histogram where there is outying data. A worst case scenario for example would be one where the whole scene is 128 grey, but it has one pixel at 16 (black) and another at 235 (white). Erik, can you add the histogram to the images in the calculator so that we can better gauge where the max and min fall as far as the bulk of the pixels?
We also haven't yet verified this model so, what I would propose is that we add some carefully selected scenes to the test pattern suite that contain large enough areas to get a consistent reading and use this data to verify the model. An example scene might be something where there is a large dark area and then a smaller brighter area large enough to get a decent measurement.
Also as far as the contrast data goes, I don't think we're quite done yet. We still don't know yet for example how well a high ANSI DLP projector fits the model (no data yet to prove the model) and there are discrepancies with the model at the low luminace portion of the curve which is a critical area for dark scenes. As Erik and I discussed I think we need to add a few more test patterns (0.5 and 2.5%) and also get the model to better predict this area of the curve (while also getting more test data with CRT's, DLP's, etc.).
It would also be nice if we could correlate a scene from the intra-scene contrast calculator to the other contrast data (and measured test pattern data). The only correlation that makes sense to me is to use average luma of a scene and convert it to the test pattern APL while realizing that the contrast that we get from the calculator is only a rough guess as to the maximum achievable contrast. I say the maximum achievable contrast, because the histogram of the pixels won't be such that all pixels are 255 and 0. If I recall though, this isn't as straightforward of a correlation as it seems. Erik, what do you think of adding a field to the table of the intra-scene calculator for each scene that attemtps to give the equivalent test pattern APL? In other words where each scene has average, max and min luma - add another field that says equivalent TP APL?
What do you guys think? Does this make sense?