ColorFoto determines the resolution at two sensitivities (ISO 100/21 ° and ISO 400/27 °) and three zoom focal lengths. The measurement method developed together with the chair for physical optics at the Cologne University of Applied Sciences uses a total of 9 siemens stars with sinusoidal modulation of the flanks. These stars are spread over the image field to capture differences from the center to the edge of the image. Subsequently, the modulation transfer function MTF is determined via the Cologne-Fit method. MTF means: What is the contrast of a recorded (photographed) structure compared to the original. Ideally, the contrast in the image corresponds to that of the original. In the case of purely optical systems, the contrast in the image is lower and decreases continuously with finer structures at higher frequencies. Digital cameras therefore increase the contrast and can thus even with rough structures in the picture achieve a higher contrast than in the subject. The test lab now determines the contrast in the image over all readable frequencies. The contrast decreases below a certain frequency below 10 percent. The corresponding frequency is now regarded as the limiting frequency, that is, the finest structure that can still be resolved. The limit frequency determines the laboratory for eight directions of each star. Over nine stars on four different levels, these values provide an accurate picture of the performance of the camera and lens.
Resolution
In our tables, you will find four resolution values for a zoom of three focal lengths: two for ISO 100 with separate details for the center of the image and the corners, and two for ISO 400 also for the center and the corners. In the point-to-point evaluation, we weight the visuals at 33 percent and the image center at 67 percent.
Object Contrast
The new method has now been tested on more than 200 cameras and results in much more reliable results than all previous methods based on a recognized physical analysis technology. This is the reason why the ISO Commission, which is also Dipl.-Ing. Dietmar Wüller from ImageEngineering is considering testing the integration into the ISO 12233 resolution standard.
Scope of the image
The optoelectronic transfer function OECF (optoelectronic conversion function) is determined by means of a circularly arranged gray scale wedge, as described in the ISO standard 14524. At the OECF, the characteristic feature of digital cameras is to convert the brightness of the image into digital values. The OECF curve is determined separately for all three color channels, red, green and blue, and provides numerous decisive information for assessing the image quality. The most important is dynamics capture, also called object contrast. It describes the maximum contrast in the recorded scene that the camera can play. If this contrast exceeds the object contrast of the camera, eg in the case of sunshine, the drawing, that is, the reproduction of details, is lost. It is said that lights emit or the shadows run. With this problem, slide films have also struggled, which bring it to an extent of about 8 blinds.
Signal-to-Noise Ratio
Negative film is more tolerant and can record contrasts of up to 12 f-stops. Newer digital cameras create up to 10 f-stops, which is sufficient in most situations - provided the exposure is correct. In order to determine the maximum detectable object contrast, the difference between two "limit" exposures is determined: a) the exposure leading to saturation (pure white area in the image file), and b) the exposure required to produce a dark Gray field, in which the image disturbances, ie the noise, are three times less than the signal. The object contrast, ie the ratio of the light reflection from the brightest to the darkest spot in the image, is expressed in densities (the logarithmic ratio of the ratio) or in the aperture levels (1 aperture ratio = approx. 0.3 densities). An object contrast of 1000: 1 corresponds to 3 densities or 10 apertures. The object contrast that a camera can detect is reduced with a higher sensitivity setting. Therefore we determine the possible contrast at ISO 100/21 ° and at ISO 400/27 °. 2 object contrasts are determined from the measured values.
Cameras with 24-bit color depth can distinguish 256 brightness levels in each of the three color channels. In the ideal case, the camera sets the lowest black to zero and the brightest white to 255. On some models, however, the brightest white is at 240. The range shows how many of the 256 brightness levels per color channel the camera actually uses, so how well Internal tonal correction works. The data are also obtained from the OECF.
With issue 11/2006, ColorFoto has introduced the new noise measurement (VN measurement) also for compact cameras. This takes the visual impression better than the old S / N measurement. A weak point remains unbunted artifacts. For digital SLR models we have used the VN measurement since issue 8/2006
Each sensor in a digital camera produces interference. Part of the interference is the "fixed pattern noise", since each pixel on a sensor has a slightly different sensitivity. When the photographer photographs a monochrome white area, the image result does not appear as a monochrome white, but with a pixel-like structure. Since this interference is specific to the camera, every camera manufacturer tries to correct the "fixed pattern noise" camera using a white calibration. Another problem is the thermal noise. Charges are not caused by incident light but by chance by temperature influences - especially at higher temperatures. In addition, the individual pixels "collect" during longer exposure times also more disturbances. The images of digital cameras are therefore better when the temperature drops and the exposure time is shorter. For this reason, the test laboratory keeps the temperature always at 23 ° C 2 ° C to create comparable conditions.
The strength of the noise also depends on the size of the sensor or each individual pixel. Smaller sensors are less sensitive to light, which means they deliver less charge with the same amount of light. Although the absolute noise is the same as a larger sensor, the signal is lower (fewer charges due to less light) and the signal-to-noise ratio is worse: the image shows a more visible noise. Furthermore, conversion to RGB images can produce color noise.
We determine the noise in 20 differently bright gray horns using the VN (visual noise) for ISO 100/21 ° and ISO 400/27 °, as the noise increases significantly with higher sensitivities and strongly depends on the brightness of the motif. All noise are captured in the test image, separated from the actual image content and evaluated. The VN value evaluates the noise according to the perception of the human eye. A decisive point is the "contrast sensitivity function" (CSF). It describes the contrast necessary for the eye to recognize the structures (here the noise artifacts) of a particular frequency. Our eye is quite insensitive to very fine structures, whereas coarser structures are perceived with a low contrast. This perception also depends on the color.
In the first step, the RGB image of the camera is transferred to the human perception-based XYZ color space and from there into the "oponent space". This "oponent space" is based on an examination of the dyes in the human eye and the processing of the resulting color signals. The image is then filtered using the "contrast sensitivity function" (CSF) frequency assuming a 40 x 60 cm print at 300 dpi. Using the XYZ color space, Image Engineering returns the image to the Luv color space, determines the default deviation for all fields, and weights the results using the following formula:
1.0000L * + 0.8520u * + 0.3230v *
The result is a curve with the visual noise (VN) applied to all 20 grayhorns of our OECF test image.
White balance
The printed average is based on the 16 average fields. We weight VN values up to 1 simply. VN values above 1 are taken to high 1.4 to take the outliers more into account. Thus, we take into account the fact that noise values of up to 1 are hardly perceived as noise, values between 1 and 2 can be seen as light noise and a really disturbing noise impression occurs at values above 3.
For an image to be neutral at different light sources, ie, gray appears as a neutral gray, the camera electronics must perform a white balance. The quality of this adjustment can be determined by the distance between the RGB values of a gray area: If the three values match, the white balance is perfect and gray appears gray. How well the white balance of the various cameras works is tested under daylight conditions. The white balance is determined by measuring the gray wedge of the OECF test chart and the RGB values of the individual fields are checked for color differences. In this case, a gray of medium brightness is rendered gray or neutral, but at the same time a lighter or darker gray shows a color tinge. Therefore, a gray wedge and not just a gray card must be recorded to determine the quality of the white balance.
Color reproduction
Whether a picture looks good can not be measured easily, but how close the colors of a photographed test chart are to the original colors can be measured. The ColorChecker DC from Gretag Macbeth, which is modeled on the colors of a natural scene, serves as testchart for this measurement. The measured values are significantly further broken down compared to other test labs. It is not only the color interval DeltaE for each color field but also the differences in color saturation, hue and brightness. These values allow conclusions to be drawn on the camera-internal color processing. The printed value indicates the mean deviation, with a stronger weighting of the critical skin and green tones.
Objective lenses are darker than the center of the image, which leads to ugly effects especially in monochrome surfaces. To measure the edge shading (vignetting), a milk glass is illuminated extremely uniformly over the special illumination, an Ulbricht sphere. For cameras with zoom lenses, the vignetting is measured at three focal lengths in the distance setting "infinity" at 1200 points in the image and output as a mean or graphic.
Another problem are curved lines at the edge of the image field. Actually straight objects transforms the lens into banana-shaped structures. Also a rectangle pattern does not remain a rectangle, but gets a barrel- or pillow-shaped structure. In order to determine the strength of this effect, the laboratory measures the ratio of the deflection of a line at the image edge in relation to the total image height. This "TV distortion" is determined as far as possible at three focal lengths.
Vignetting
Distortion
Maximum Imaging Scale
Switch-on time
Trip delay
With many digital cameras, macro recordings can be made: in order to determine the size of the smallest imageable area, we photograph a scale with the optimum focal length from the shortest distance at which the camera is still focused. Specifies the width x height of the smallest imageable area.
Fast response to interesting motifs and corresponding snapshots are not possible if the digital camera takes 30 seconds until it is ready for use. The lab sets the time until a camera is ready for use.
To help you know how fast or slow a camera is responding, we measure the time that passes from pressing the shutter button to the recording to 1/100 sec. A LED array with a total of 100 LEDs serves as test motif. First, the camera is focused on an object at least 100 m away and then the LED array is located at a distance of 1.5 m away from the AF-meter. The first LED flashes at the same time as the pressure on the trigger, and the second, third to the 100th LED flashes at a distance of 1/100 s. Then a look at the recording, at which LED triggered the camera. The time required can thus be determined by the LED that is just flashing. In this triggering delay the autofocus time as well as various internal time-consuming calculations fall. For zoom lenses, the focal length is adjusted so that the camera image-fills the test array.
How fast is a camera ready for use after shooting? Our laboratory determines the possible number of serial images per second for a typical JPEG image with maximum resolution, lowest and medium compression. Special serial image functions for lower resolutions are not taken into account. In addition, we determine the maximum image count per series.
Sometimes CCD sensors in digital cameras have defects, which should be corrected if possible in the camera. We are checking whether the case is black. The procedure is as follows: with the cap, two exposures of 4 and 1/15 s are taken each time after switching on the camera. All pixels in this black image that have a digital value greater than 20 are considered defective.
While the quality of the built-in zoom plays a decisive role in compact cameras, the SLR test should run as objectively independent as possible. Finally you can choose different zooms and fixed focal lengths for the respective SLR. For the measurement of the resolution we therefore use a high-resolution macro and determine the resolution exclusively in the center of the image. The specified value corresponds to the maximum achievable resolution, which is therefore not achieved by most lenses.
Image sequence times
Flashy pixel groups
Differing test procedure for digital SLR cameras
No comments:
Post a Comment