Adaptation study die general-purpose image measurement manner

摘要

PURPOSE:To measure a picture adaptively to various kinds of applications by performing statistical feature extraction with a learning function by using a multivariate analyzing method according to an extracted initial feature. CONSTITUTION:Picture measurement can be considered to be the problem of feature extraction from the picture. As its basic condition, there are three items of 1) variability in respect of parallel displacement, 2) additivity in respect of the picture, 3) adaptive learnability. In geometrical feature extraction F1, a large number of general and basic features to satisfy the conditions 1 and 2 are extracted as the initial features for all screens. In the statistical feature extraction F1, the features optimum for various kinds of the application are extracted adaptively by learning by linear-coupling the initial features. Thus, the measurement based on two stages of the feature extraction satisfies the conditions 1 to 3 of the picture measurement, and the picture measurement can be performed adaptively for various kinds of the applications by the learning of high speed.