Bivariate Estimator for cDNA Microarray Images Using Complex Wavelets

Removal of noise is an essential step in the preprocessing of microarray images for obtaining betterquality gene expression measurements. Wavelet-based methods for denoising of images are very successful. However, for cDNA microarray images, existing methods are not as efficient because they fail to take into account linear dependencies that exist between wavelet coefficients of the red and green channel images. To address this issue, a bivariate MAP estimator is proposed in the complex wavelet transform (CWT) domain that jointly estimates wavelet coefficients in the two channels. The CWT is preferable to the traditional discrete wavelet transform for denoising of microarray images owing to its good directional selectivity and shift-invariance properties. Both properties ensure better detection of edges in the spots. The proposed method is compared with other locallyadaptive CWT-based denoising techniques using simulation experiments. Results show that our method achieves improved noise reduction performance as compared to others in terms of the mean squared error.