Resolution

In the first step, we determine the resolution with a black and white Siemens star. Its edges are not sharp, but instead softly modulated equal to a sine curve, so that the re-sharpening of the smart phone is only applied moderately. The boundary of the resolution is reached when the image‘s contrast is reduced to 10% of the initial value (MTF10). Each Siemens star is devided into eight segments, which are evaluated separately and then averaged. Thus, the result is independent from the direction of view. On our test chart, you find seven Siemens stars with high contrast (1) and three Siemens stars with low contrast (2). As cameras utilize contrast-dependent algorithms, it is important to represent this in the test.

Optical Trimming

As the Siemens stars are located at three different distances to the center of the image, we can identify the trimming of the resulution. This is first of all a property of the optical lens. Under normal circumstances, the resolution drops more or less at the edges, depending on the quality of the optical lens. But signal processing again plays a role in this as well. Because with re-sharpening and edge boosting, the resolution can be influenced overall. This typically leads to higher resolutions in the corners as in the image center. But when a manufacturer applys these techniques too aggressively, the result is an unnatural, too hard and artificial visual impression.

Centering

Especially in smart phones, the optical lens is not always perfectly adjusted directly in front of the sensor. Also, transportation of the device can lead to misalignments. For this reason, we compare the resolution of the four Siemens stars in the image‘s corners. For badly centered lenses, we measure differences of up to several hundred line pairs per image height.

Detail

Transversal from the Dead Leaves field with high contrast (3 in test chart shown at the top of this page), the test chart contains a Dead Leaves field with low contrast (4). On these coloured structures, we also determine the resolution for both high as well as low contrast textures. After all, the world is not only colourful, but also full of different contrasts.
On the Dead Leaves fields we measure the resolution of the coloured structures against a threshold contrast value of 50%. So the boundary of the Dead Leaves resolution is reached when the image contrast drops to 50% of the initial contrast (MTF50). Thus, the low contrast structures for example deliver a measurement value for the preservation of fine textures without maximum contrast in the image.

Artefacts

As part of image optimization, a camera software can add artefacts as additional and hence erroneous structures into the image. For the Dead Leaves fields, we therefore compare the camera‘s image with the original picture on the test chart. This way, the test software recognizes which structures were preserved in the image at which level of contrast – and which were newly added as artefacts. This leads to a clean DL
value and at the same time delivers a measurement value for artefacts, again in relation to high contrast and low contrast textures.

Noise

Our test procedure makes use of the noise evaluation VN which is adapted to visual perception. High VN values designate much noise. In addition to the grey fields (5 in test chart shown at the top of this page) we also consider the noise in the colour fields (6). Furthermore, noise is dependent from brightness, which is also registered.

Edge Accentuation

All digital cameras optimize edge accentuation, in order to make the image look sharper and more rich in detail. This enhancement improves the values for resolution and Dead Leaves, which is absolutely reasonable if used modestly. Without any edge accentuation, the images appear contourless. However when this is exaggerated, the resulting image will look damaged.
If a camera emphasizes an edge, the recorded flattened rectangular curve will not result in the ideal rectangular curve of the original, but in an amplified one with slight over and under shoots. This is desirable in moderate use, but often tends to be exaggerated. In the resulting image, ugly parallel lines then accompany the edges, which can be either light or dark.
On our test chart, the edges (7) are positioned at two distances in relation to the image height – always as a pair with high and low contrast, and always aligned both horizontally as well as vertically. A resharpening depending on direction is not unusual in cameras, and thus can be determined in our testing. The calculation basis for edge effects is the surface below the over and under shoots. We use the edge values thus calculated for high and low contrast in order to evaluate the according Dead Leaves and resolution results.

Colour Representation

Testlab not only determines the colour interval DeltaE for each colour field (6 in test chart shown at the top of this page), but also the difference in colour saturation, in colour hue and in brightness. The published value states the mean deviation, DeltaE.

DISTORTION

Contorted lines at the image‘s borders are particularly well known from super wide angle lenses. They result in representing a straight wall as a curved shape. For wide angle and tele lenses, the camera software eliminates these errors more or less completely from the images, which works very well in most cases. Still, in the corners circles can sometimes be transformed into small, diagonal lines. Depending on the manufacturer, the distortion correction can be switched on or off for super wide angle cameras. The very heavy corrections which are often necessary for these lenses can lead to distortions in the corners, reduced resolution, a rise in noise etc. Testlab determines the distortions still present in the image with the help of register marks.

Vignetting

Almost all wide angle images exhibit more or less strongly shaded image corners when uncorrected. In addition to dimming being caused by lens design, the cos4 law is to “blame“. It describes the loss of brightness in the corners depending on the image angle. Manufacturers try to counteract this with optimized computing and a software-based brightening of the corners internally in the camera. However, this can cause noise to increase in the corners as well. Thus, our measurement compares the brightness of the test chart‘s background between the image and the original.

Interdependence of Sharpness, Contrast and Resolution

Sharpness, contrast and resolution are different properties of an image, which are closely connected. Our measurements determine the resolution of fine details and consider the contrast, but not the sharpness – which would be the steepness of an edge. Regarding resolution, the question is: How fine can lines become in order to still stay distinguishable and not vanish in a uniform grey? This resolution threshold (here measured on Siemens stars – 1 and 2 in test chart shown at the top of this page) is determined for a contrast of 10% of the original value and then converted into the number of line pairs which would fit into the height of the image. According to this definition, a structure still counts as resolved, when its contrast has dropped to 10% of the initial value (MTF 10). This treshold frequency is determined for high-contrast and low-contrast textures. The determination of resolution is thus based on contrast. For the coloured Dead Leaves fields (3 and 4) we assess the MTF 50 – which means that we look for the treshold frequency at which the image‘s contrast drops to 50 percent, again for high-contrast and low-contrast textures. Both put together ultimately provides a good measure for resolution and detail.

The following video provides additional insights into the testing procedure and the facilities of Testlab.