Impulse noise removal in two-dimensional electrophoresis images based on dome recognition

Qiaofeng Ou,Huisheng Zhang,Lixin Li,Qiaofeng Ou,Bangshu Xiong
DOI: https://doi.org/10.1109/BMEI.2015.7401545
2015-01-01
Abstract:Two-dimensional gel electrophoresis (2DE) images are often corrupted by impulse noise in broad sense (including various artifacts, such as fingerprints, hairs, gel cracks, strips, water stains, dust and so on). In this paper, we propose a novel adaptive directional filter (ADF) to remove impulse noise completely and accurately. Firstly, we detect the impulse noise based on the observation that the directional derivative distributions around the protein spot centers and noise pixels are very different. Then, according to the bell-shaped three-dimensional appearance of protein spots, the greatest and smallest gradient directions were used to restore the noised pixels. Lastly, the pixels adjacent to the modified pixels are iteratively detected and restored until no impulse noise is detected. Experimental results on the synthetic and real 2DE images show that the proposed method performs much better than the previous approaches.
What problem does this paper attempt to address?