[en] The paper focuses on the quality of predicting the damage characteristics of the loaded engineering structure by subjecting section images of cores supposedly drawn from the concrete of the structure to quantitative image analysis by sweeping test lines. Automation of this data acquisition stage is shown, generally leading to information that is biased to an unknown degree. This is accomplished mathematically and graphically according to Underwood. When the Stroeven/Saltikov (S/S)-concept of a partially linear-planar model for damage is accepted and (four-connexity) digitization is accomplished in a direction adjusted to the prevailing orientation direction of the cracks, the paper shows that orthogonal observations in vertical sections are sufficient for the unbiased assessment of total crack length per unit of area (2D), or specific crack surface area (3D), also when digitized images are employed. This is possible in situations where (uniaxial) compressive or tensile stresses dominate. The crack orientation distribution is however always biased when determined on digitized images in an automated set up when pixel directions are not compensated for, such as by the quantified S/S-concept of damage.