Reproducible display sparkle measurements

An autofocus algorithm tries to optimize the focus to get a sharp image, and in our case, this is the pixel layer. If you use an autofocus technique to go to the pixel layer, then you cannot be sure where exactly in the depth of focus you are located. And this unsure point, where you are within the depth of focus is the main problem that you have when using an autofocus technique because the autofocus also relies on the signals you get on the sensor levels, and within the depth of focus it is the same signal - the same is also true for manual focus. As many of the current systems have a fixed F-Number, this is the most effective way to overcome this.

That is the advantage of the Fourier transformation approach. The filter in the frequency approach does not only filter the periodic component, but also the low-frequency components. The influence of the backlight is a low-frequency component, so we completely removed it. Also, the local evaluation would remove its influence because we are always dividing by the local mean of this area to get the local sparkle value. So, we do not have an influence on the backlight or of low-frequency components.

The method can be used for large curved displays as well because you do not have to measure the sparkle on the complete display. It is fine to only measure a small portion in the middle. If the display is not too curved, it should not be an issue. If you measure a display with a higher curvature, you should probably choose a setup with a larger depth of focus.

One measurement was made for a 50% luminance, but there was no noticeable difference in the results. It would not make much sense to measure the sparkle for a lower luminance.
The measurements are color-sensitive. In this setup, we measured the sparkle for red, green, blue, and white. Although we received the highest sparkle values for red and blue images, we decided to use green. This is because most existent methods use green, and humans perceive sparkle best when using green.

Thanks to our filter in the frequency domain, all low frequencies are being filtered out. If there is such an effect, it will not be measured directly.

It depends on the PPI range. We compared displays with 224 PPI and 183 PPI, where it did not matter whether you keep the magnification or reproduction scale constant. 

Our partners from VW have set the boundary to approx. 5% for the RPS of 2.45.

The algorithm, which will be part of the specification, is removing the frequencies in the FFT.

We tested the sparkle effect for several lenses and camera combinations, looked at the sparkle matrix, and calculated the field angle. The effect is stronger for the 16 mm lens because of the larger field angle.

The measurement itself is only a one-shot measurement. The alignment depends on whether you are using an LMK Position, where it takes about 40 seconds or another set-up. In the latter case, the alignments must be in according to BlackMURA. Another part of the measurement setup is finding the correct focus, for that you need to take several pictures to find the correct focus.
If you can ensure that the display or the anti-glare layer will always be placed approximately in the same position, the following measurements will be one-shots. And as sparkle is usually measured using a 100% green display (green ensures better visibility of the sparkle), you also have a short integration time.

Yes. You will need the updated software, the display sparkle add-on, as well as a translation stage. This can be one you already have, but you can also get one from us.

We will specify this. We tried to make the value as low as possible. A lower reproduction scale means that you get closer to a BlackMURA compliant setup.

With a sparkle matrix, you can directly measure the influence because the camera has different observation angles. This effect can be observed in this picture, this was a 16mm lens. There is a very high sensitivity of the measurement field angle. We observed this for all the other glasses we tested. This effect is affecting visibility. You can also try to calculate another merit in this image if you are interested in specific angles. However, for a single value, the median is better, and a limitation is required. Even so, the effect itself can be interesting.

We did not measure this effect. If you want to do this, you will need to adjust the focus position, and scan in another area, where you should try to find something. However, it will be hard to distinguish between sparkle values from the glass and sparkle values from behind.

In general, the calculation is based on image processing, however, those are a bit complicated, but they will be part of the measurement specification. This is particularly important for equipment manufacturers, who will implement this technique.

The reproducibility is lower in this case, therefore, we are talking about the same range. There was not much of a distance from a perceived point of view. Also, a different measurement field angle was used.

The display used is an automotive display, which already had layers, so also a sparkle.
But even if the display were perfect, you would not get 0% or close to 0% because of a small sparkle offset. The filtering frequency domain introduces other effects, and these other effects also create sparkle.

Sparkle is a phenomenon that can be observed by the human eye, so you should use luminance measurements at the beginning to get a meaningful value for the sparkle, and with this, traceability is associated, so you need a calibrated system to perform this kind of measurement.

That depends on your measurement setup, as well as on the handling issues addressed in this online seminar. There is a strong change in the sparkle. If you have small tiny defects or a large over-angle effect, then you have big maxima.

Regarding the reproduction scale, you will have to agree on a value.
We will write a value for that in the VW specification. The field angle is not that critical regarding this value. You can do the calculations and ignore the field angle.
You should only use the area which covers the field angle you are interested in if you use median-based local evaluation, and only for the last evaluation.
The field should not be too small, it doesn't make sense to only measure a few pixels, but in a practical application with 16 mm lenses and the reproduction scale that we talked about, this is not really a limitation. This is more about the evaluation area, which you must consider. For the reproduction scale, we try to use small values. It is still under research how small we can get. The smaller we get, the closer we get to the existent BlackMURA set-up. Until now, the lowest value that we found is 2.1, but we are currently testing values that are lower than 2.

We have not made a physio psychological study. As of now, there is no real study for the level of acceptable sparkle, only the expert study showed in the online seminar. We have the values low, medium, and high. This range was provided to us by our partners, the sparkle was clearly visible. Our partners find that an acceptable level is in the range of 5%. If it is more than 5%, they would not use the cover or glass because there would be too much sparkle. In other cases, it would be fine because other effects of the anti-glare layers should be considered.

The parameters to be shared are the maximum measurement field angle and the reproduction scale. The reproduction scale will probably be specified in the specification, and then you would only have to share the maximum field angle.

Yes, that is the plan. The details regarding how to get the frequency filter were skipped in the online seminar. It will however be described in detail in the specification, such that anyone who wants to implement the algorithm, e.g. an equipment manufacturer, can do this based on the specification. BlackMURA is our layout for that.

We did not use a telecentric lens, we only used the standard 16mm, 25mm, and 50 mm lenses. The reason for that is those are the lenses that are currently used for BlackMURA set-ups and measurements.