Detection of local objects on digital microscopic images

The problem of detection of local objects, small in comparison with the image size is considered. It is shown that two-stage solution of the problem is necessary - preliminary processing of the image and the search for objects. Preliminary processing includes peaking filtration of the image by Tikhonov or Wiener filter, intended for discerning of conglutinated objects and for their normalization. Besides at this stage the background component of the image is eliminated. A search algorithm using formation of a threshold surface of detection is described. It is experimentally shown, that the detection efficiency exceeds 90 %, and the processing time is less than 60 ms for the image size 256 × 256.