The correct exposure is an important aspect for evaluating test photos. The measurement results of correctly exposed, over and underexposed shots can differ from each other, especially with camera generated JPG images. The reason for this lies in the camera’s internal process system and the optimization of the data itself.
While the data of RAW images are normally untouched by additional internal camera processes, different in-camera processing steps take place when producing a JPG file. In addition to the assignment of the color and the brightness adjustment, other processes are often used for image optimization, like noise reduction, contrast adjustment or sharpening. The processing procedure differs mostly depending on the manufacturer and model. The photographer also influences the optimization with his chosen camera settings (white balance, sharpening, for example). The individual steps, the sequences and the type of the procedure are mostly kept as a secret by manufacturers and are difficult to retrace from a generated data.
Correct exposure for comparable results
Why do these steps play a role in analyzing test images that are taken with lensTESTER? The crux of the issue is linearization. The signal of a JPG-image behaves nonlinearly with respect to the brightness. If a test image is analyzed with this nonlinear behavior, the result would be falsified measurements relative to the brightness.
For this reason the nonlinear signal has to be recalculated into a linear signal (see Linear/non-linear/linearization). The characteristic curves of the profile are used for linearization. Because the steps of the internal process are unclear, the camera’s internal optimization could still influence the results. For example, sharpening, which is carried out mainly by an increase in contrast especially at the edges.
For comparison, we contrast JPG and RAW images, taken with different shutter speed settings. In this case, the RAW images were developed in a nearly uncompressed 8-bit TIFF-format (sRGB).
The different shutter speeds have barely any influence on the calculated results of the TIFF images. The camera generated JPG images show instead at lower spatial frequencies, a variations in contrast. The raising curves indicate a camera internal sharpening, which took place during image processing and changes with the exposure times.
Because a recording system’s correctly exposed, overexposed and underexposed JPG test images (with the same aperture, ISO and focal length settings) are hardly comparable, we must assume a correct exposure. Only under this condition is a meaningful comparison of the test results possible. Because of this, the statistic average is calculated only for images, which were taken under this requirement.
To ensure that correct exposure is present, every individual test photo has to pass a “Pre-Check”. During this process, the images are checked for correct exposure, illumination and other aspects. Over and underexposed photos are not generally excluded from the analysis. A certain range is granted for evaluation. Therefore, slightly over and underexposed test images are permitted. However, because of the influence of the internal image processing system, the results from those photos do not flow into the calculation of the statistic average.
In the course of linearization a so-called normalization takes place. This is necessary to adjust the calculated modulation (contrast value) according to the image taken. With normalization, the calculated modulation will be standardized over the contrast given by the recorded image, using the black and white reference circles surrounding a Siemens star. In connection to a correction factor, which is implied for calculating the contrast values, thus the sharpening in dark areas can be detected.