Display screne image recognition and bounds of number of steps of an algorithm in the frameworks of logic-objective recognition system

The paper is devoted to the proof of upper bounds of steps of logic-objectiv algorithms for recognition of a complicated image situated on a display screne. It is proved that the problem of separation and recognition of an etalon object from a complicated scene has a polynomial algorithm. The problem of separation and recognition of an object from a class the description of which contains only distinctive attributes of this class belongs to NP. To decrease the algorithm number of steps a notion of “fuzzy image” is introduced. The problem of invariant (under rescaling) image recognition is regarded.