Effect of Smoothing Filters on Time-Domain Waveform Image Segmentation

Abstract

In this paper, we studied the effect of smoothing filters on the segmentation of Time-domain wave form image (TDWI). The positive and negative impacts of global and localized smoothing operator shave analyzed and forwarded our conclusions. We used two class of images for the experiment, ordinary Lena test images, and Time-domain waveform images. All images are first converted to the grayscale then segmented without smoothing and after smoothing with both global and localized filters, the results are compared using a visual comparison, Peak Signal to Noise Ratio (PSNR) and Means Squared Error (MSE).There are lots of studies that generally recommend using general smoothing operators for better output of image segmentation on various applications. But we realized that smoothing TDWIs needs special attention on selecting the type of smoothing operators to use. Using the global smoothing operators before segmentation has shown a deleterious impact on the pertinent information of the segmented waveform. The experiments proved that the effect is drastic and fatal to the entropy of waveform because the Region of Interest (ROI) on the image is a tiny line structure with a width consisting only a few pixels. As a result of the large mask window of the global operator, the central pixel intensity may average out to the other region which inturn leads to change in shape of the wave form. This is significant to change the pertinent information. But using localized smoothing filters have shown outperforming results. Thus we propose to avoid using global smoothing operators during segmentation of TDWIs. But localized smoothing filters with simple basic image enhancement techniques are strongly recommended.