Challenges in Automated Inspection of Display DevicesAutomated quality inspection is done on many areas of industry but it is especially demanding for displays, when display operation is under inspection. This article discusses some of the problems you may encounter when implementing automated display inspection. Solving these problems formed the base for a system design work, which Atostek had carried out for a customer.
The main idea in automated display inspection is to automatically check whether the desired image is shown on the display, and automatically recognise artifacts from an acquired picture. Here the picture is the image that has been digitally recorded with a camera from the display (Figure 1.). The automated display inspection system consists of two main parts; picture acquisition and analysis (Figure 2.). Even though it sounds easy, simply taking a picture of the display is quite challenging and even more so is the analysis that is needed to make sure that the images are correctly displayed.
The display inspection process starts with the picture acquisition. At this stage picture quality is most important, since bad quality will make the comparison hard or even impossible. There are several challenges that need to be overcome at this stage. Two of them are discussed here:
Aliasing is a well known phenomenon in digital signal processing. When an analog signal is digitised with a sample frequency that is less than twice the highest frequency content of the signal, an aliasing effect will appear. The high frequencies will mirror themselves to a lower frequency band. This mirroring of the signal is irreversible and cannot be removed after the digitalisation of the signal.
The aliasing effect is known as Moiré effect in digital image processing (Figure 3.). This is one of the first problems encountered when taking a picture of a display. High resolution of the display means the acquired picture contains high frequencies. This is specially pronounced when taking a picture of a colour display, since all the pixels consist of three subpixels (Figure 4.), which will considerably increase the high frequency content.
White light has a colour temperature, which is measured in degrees Kelvin (K). For example, daylight has a colour temperature of 6500 K and analog television of around 9300 K. The colour temperature of an LCD display depends on the backlight of the display and the colour filters used in subpixels. These parameters change from one display model to another. Differences in the colour temperature make it hard to recognise different colours from a display, since the colours are not pure. For example, white can be seen as yellowish or bluish (Figure 5).
After the picture has been successfully acquired it must be analysed. There are several challenges in the picture analysis, two of which are covered here:
Because of colour temperature, camera technology, picture resizing, and several other reasons the colours of the original image and the picture do not match. It is fairly easy to compare gray scale images, but it is several times harder to do the same for colour images. For example, a yellow colour that human easily recognises to be yellow, might be categorised as orange or green by a computer.
In order to compare the original image and the acquired picture, the picture has to be resized to equal size with the image. This is not straightforward, since the picture does not contain just the pixels, but also the black spaces between them. Resizing algorithm has to be designed carefully, so that it removes the black spaces but does not lose small details.
The challenges discussed here represent some, but not all of the problems to be solved in a system design project like this. One of Atostek’s specialties is the design and implementation of advanced machine vision systems, such as described above.