Content-based leukocyte image retrieval ensembling quaternion fourier transform and gabor-wavelet features

This paper addresses the development of content-based medical image retrieval system for automated leukocyte recognition from microscopic images of peripheral blood smears. The proposed methodology ensembles Quaternion Fourier transform based segmentation and Gabor-wavelet describing textural features of leukocytes. This technique is computationally efficient because it captures local image information. The meaningful local regions of an image are segmented automatically without prior information. Gabor-wavelet based textural features are extracted from segmented leukocytes for image retrieval. Experimental results show 73.6% precision and 77.8% recall with respect to normal query image of leukocyte and 76.1% precision and 78.4% recall with respect CML query image.