Currently, in order to increase the speed of preparation of the ultrasound control protocol and reduce the influence of the human factor, reflector recognition (classification) systems based on artificial neural networks are being actively developed. For their more efficient operation, the images of the reflectors must be processed in order to increase the signal-to-noise ratio of the image and its segmentation (clustering). One of the segmentation methods consists in image processing with an adaptive anisotropic diffuse filter, which is used for processing optical images. In model experiments, the effectiveness of using this texture filter for segmentation of images of reflectors reconstructed from echo signals measured using antenna arrays has been demonstrated.