Wavelet Denoising Algorithm Based on NDOA Compressed Sensing for Fluorescence Image of Microarray

Abstract

A microarray can be easily used for quantitatively analyzing the expression levels of DNA genes. Yet, the noises introduced during the application will greatly affect the accuracy of DNA sequence detection. How to reduce the noise constitutes a challenging problem in microarray analysis. Especially, due to the weak fluorescence response, the image of microarray contains difficulties of the low peak-signal-to-noise ratio (PSNR) and luminance contrast. To solve the problem that the wavelet threshold denoising method has poor effective on low PSNR image, a wavelet denoising approach based on compression sensing (CS) optimized by the neural dynamics optimization algorithm (NDOA) is proposed, which preferably solves the denoising difficulties of noise pollution in the microarray image. Under the condition of Gaussian random observation matrix, the effectiveness of NDOA-optimized wavelet denoising based on CS gets better work than the orthogonal matching pursuit and its improved algorithms. The experimental results indicate that the expected wavelet coefficients of the noiseless image have been reconstructed with higher quality. When the compression sampling rate for microarray image is 0.875, the PSNR of the NDOA-optimized wavelet denoising algorithm based on CS is increased about 9 dB, and the root mean squared error is reduced obviously too, in comparison with the wavelet soft-threshold denoising method. It shows that the NDOA-optimized method improves the performance of the classical wavelet threshold denoising.