Characterization of biomedical images using Walsh filters followed by convolution

TLDR: In this paper, a directional and selective edge highlighting using Walsh filters has been proposed to distinguish between a normal cell and a diseased cell using this selective edge enhancement followed by arc convolutions.

Abstract: In edge enhancement, edges in all directions are usually evenly highlighted in image to demarcate the shape of biological cells apart from other applications. Many diseased cells and parasites deviate from the regularity of a normal cell in terms of shape with deformity increasing with stages of the disease. In this proceeding, a directional / selective edge highlighting using Walsh filters has been proposed. A differentiation between a normal cell and a diseased cell using this selective edge enhancement followed by arc convolutions has been demonstrated. Here, the sample image in spectral domain is multiplied with the Walsh filter before inverse Fourier transform is taken to obtain selectively highlighted edges. Subsequently, arcs of different orientations are convolved with the edge enhanced image. Resulting convolution peaks are isolated from the background by proper thresholding and counted using morphological techniques. For a symmetric cell, convolution with arcs at different orientations gives same number of peaks whereas for asymmetric cells different number of peaks will be obtained. Here, proper thresholding to eliminate lower peaks and selective edge enhancement to highlight or suppress (for example suppressing diameter in semi-circle) certain edges is very important. Also, the positions of the convolution peaks depending upon the arc orientation give edge direction. Spatial Light Modulator (SLM) is usable for the edge enhanced image generation. Additionally, Hilbert transform for edge enhancement has also been used.