Abstract: The invention provides machine vision systems and methods for morphological transformation of a source image, e.g., with non-uniform offsets. According to these, pixels from selected “neighborhoods,” or regions, of the source image are rearranged, or unfolded, into respective columns of an intermediate image. For each column in the intermediate image, the method identifies a pixel intensity value of selected rank. Once the pixel intensity value of selected rank is identified for each column, it is retained in a transformation image, e.g., at a location corresponding to center of the respective neighborhood. The foregoing steps are repeated, e.g., along strips or bands, for each neighborhood in the source image.

Abstract: The invention provides machine vision systems for morphological transformation of a source image e.g., adapted for use with zero or uniform offsets. The systems have application, for example, in image erosion or dilation of images of varied image quality. Systems according to the invention compare each pixel in a first row (or line) of the image with a corresponding pixel in a second row of the image. In each comparison, the system identifies the pixel intensity of selected rank. Where the system effects a dilation-type transformation, for example, the comparison seeks a maximum of the compared pixels. For an erosion-type transformation, the comparison seeks a minimum. The pixel intensity value of selected rank (e.g., minimum or maximum) determined from each comparison is retained and, in turn, compared with the intensity values of the corresponding pixels in every other row in the image.

Abstract: In an image retrieval system, a database with a large number of images is searched to find one or more images meeting the specification of a user. This specification is given in the form of a query image. The system determines the similarity between the query image and a particular image from the database by comparing the color histograms of the two images. The histograms are treated as statistical distributions and the similarity is determined on the basis of an information theoretic measure of the distributions. In a first embodiment, the similarity is determined using the Kullback informational divergence of the two histograms. In a second embodiment, the similarity is based on the entropy of the distribution of similarity coefficients of the two histograms is used.