Automatic detection of spots in biological images by a wavelet-based selective filtering technique

We present a method whereby the extraction of spots from grey-level biological images is accomplished by globally transforming the image and by locally and adaptatively modifying the transformed image. The method is based upon selectively filtering an undecimated wavelet decomposition of the image through the use of wavelet coefficient thresholding and correlation. Since spots are big local discontinuities in the image, the proposed strategy is to trace and to reinforce their wavelet signature across a number of resolution levels and to use the resulting information to accomplish recognition. Results are presented for the analysis of typical immunofluorescence and colloidal particles immunolabelled images.