A great number of image processing problems can be stated as approximation problems. A class of approximates can be selected optionally, however, it is more effective to select by function, which parameters have a simple physical or technical interpretation. To verify the theoretical model of the effect or of the research object measurement of a difference between theoretical data (or hypothesis) and experimental data only is useful. The measurement problem can be stated as follows: (1) We have a hypotheses about the class of object function. (2) We have some nonoptical quantitative data. (3) We have an image connected with the object. (4) We have a hypothesis about errors of all data sets. (5) We are looking for an object function fitted to the experimentally determined image. The problem can be solved by use of the Tichonov regularization method, which is a good idea for stable solution of inverse problems.