Method and device for the characterization and localization in real time of singular features in a digitalized image, notably for the recognition of shapes in a scene analysis processing operation

摘要

The singularities constituting the characteristic points of the image are points of contour with sharp curvature such as corners or regions of small size, the image being formed by a two-dimensional frame of pixels each exhibiting a defined grey level. The device of the invention comprises: differentiating means (10) approximating, for each pixel to be analysed, the partial, second-order derivatives of the grey level function of the image (F(x,y)); characterising means (23, 30), determining on the basis of these derivatives an attribute ( lambda 2) representative of the characteristic sought and assigning this attribute to the pixel thus analysed; and localiser means, searching among the pixels of the image thus analysed for the points maximising the said attribute. Very advantageously, the differentiating means (10) comprise convolving means carrying out a convolution of the grey level function of the image (F(x,y)) with functions corresponding to the partial, second-order derivatives of a regularisation function, this regularisation function making it possible to attenuate the noise present in the image. Preferably, the attribute chosen is the second eigenvalue of the matrix of the second derivatives, and the regularisation function is a Gaussian function, especially a centred Gaussian function. IMAGE