Abstract
A computer model of adaptive segmentation of two-dimensional visual objects was developed, based on neurophysiological and psychophysiological principles. The model imitates several stages of visual information processing. At the first stage, a preliminary assessment of the image is performed using a brightness pattern analyzer. At the second stage, control parameters are formed on the bases of the initial assessment. A defined control vector is synthesized for each type of starting image, which allows adaptive processing; this initiates two parallel mechanisms of primary image description: one contains the outlines of images and the sharp boundaries between their fragments, and the second contains areas of uniform intensity. The control parameter vector is applied to these two descriptions to analyze their brightnesses and spatial characteristics, and this is used as the basis for forming the regime for subsequent processing, which includes a set of processing operators whose parameters are tuned for each fragment of the image. The primary phasic and tonic descriptions are then used to extract individual fragments of the image (i.e., discrimination of figures from the background) and to form the final presentation. The resulting descriptions complement each other, creating a basis for quantitative measurements of image characteristics and allowing various signs needed for image classification to be formed. The computer program for adaptive image segmentation was tested using a large number of different two-dimensional half-tone objects, the purposes of these exercises including segmentation and measurement of objects in morphometric and cytometric studies.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have