The article studies the adequacy of recognition methods on different classes of images. The recognition methods include methods using wavelet transform and Fourier transform, as well as weighted additive convolution of the evaluation criteria corresponding to these methods. Identification of the weights of the evaluation criteria as part of the additive convolution is carried out on two sets of one-dimensional signals composed after linearization of the corresponding image families formed by uniformly shifting the standard image to the right horizontally. The image of the state emblem of the Republic of Azerbaijan is chosen as such standard. The derivation of a numerical evaluation criterion for comparing the declared recognition methods is carried out empirically on the basis of artificially generated image families reflecting the state emblems of separately selected countries. Two numerical criteria for assessing the accuracy of the declared recognition methods on specific image families are experimentally established and verified.