A Stable Comparison Model Of Denoising Methods Based On Wavelet Transform

The surface plasmon resonance imaging signal (SPRI) is an image signalobtained by surface plasmon resonance sensor. Accurate SPRI detection in practice islimited by several kinds of noise, including the noise in detection electronics and thefluctuations of the light source. To improve resolution of the SPRI technique, weshould use denoising methods to process the SPRI signals. The noise reductionperformance of the methods is often evaluated by calculating the signal-to-noise ratio(SNR) between the original signals and the denoised ones. It seems that the methodproducing larger SNR can suppress noise better. However, if the reduced noiseapproaches zero in practice, SNR can also reach a large value. In this case, we can’tget the expected denoising effect. When one has to choose an optimized denoisingmethod, how to use the SNR calculated appropriately should be considered. If thesignal wanted can be predicted before the denoising step, we can calculate the wantedSNR and choose the denoising methods producing most approximated SNR to thewanted SNR as the optimized method. Thus in practice, before evaluating thedenoising methods for a specific signal, we should simply estimate the theoreticalsignal and the exact value of the SNR between the measured signal and its theoreticalsignal. For molecular binding experiments, if we want to get kinetic curves from SPRIsignals, we can either extract the curves directly from the signals and process thecurves by signal processing methods to get the denoised curves, or process the signalsas images by image processing methods and extract the denoised curves. We find themethods based on wavelet transform (WT) have been applied to both signalprocessing and image processing, thus we will take WT denoising methods of imageprocessing and WT denoising methods of signal processing as denoising methodexamples in the demonstration. In this condition, we need to establish a comparison model to get the optimized denoising method.In this paper, we propose a stable comparison model stable comparison modelof denoising methods for surface plasmon resonance imaging signal. This modelallows one to get the optimized denoising method. We can use the method to suppressthe noise in SPRI signals effectively. Next, we will prove the feasibility of the modelin a statistical way. We first demonstrate the model by getting the optimizeddenoising method. To make the demonstration more straightforward, we calculate agroup of kinetic curves for our designed biomolecular interactions according to ourcurrent SPRI instrument and construct SPRI signals based on them with randomSNRs by adding different noise intensity. Since the SNRs in the demonstration are ofrandom values, it helps to test the robust properties of our model. We will use themodel to get the optimized denoising method and calculate the mean square errorbetween the calculated kinetic curves and the denoised curves from the denoisingmethods. In this calculation we find the mean square error between the theoreticalcurves and the denoised kinetic curves from the optimized method approaches zero,which means it is the most approximated signal to the calculated kinetic curves.Application of the optimized denoising method obtained from the model to SPRIsignals helps to improve the resolution of SPRI instrument.