Experiemental Task for Digital Imaging & PACS

The Hong Kong Polytechnic University

Course:  Digital imaging and PACS

Subject Code: HTI 5720

Experimental Tasks Report:    Digital Imaging and PACS Experimental Tasks

Submitted to:      Dr. Lawrence WC Chan

Student Name:        Cheng Pak Kwan

Student ID:                10673287G               

Submission date:   19^th April, 2011

Study Title

Compare between Linear filter (Average Filter) to Nonlinear filters (Median Filter) by using MatLab.

Research Question and Objectives

How effect the image quality if apply different types of linear and nonlinear noise filters are used in abdomen CT?

u  To investigate the change in image quality if different types of linear and non-linear noise filters are used in the image processing of abdominal CT images

u  To evaluate which block size is the optimum matrix filter size can suppress more image noise to improve the image quality in abdominal CT images

Experimental Design and Plan

Image filters and Computer Aided Diagnosis (CAD) tools are necessary tools for clinician to diagnosis the digital image from viewing workstation. If they applied the suitable image filter, which will greatly improved images quality. By means of this approach using appropriate digital filter, which not only suppress the noise, but also can reduce some unnecessary patient absorption dose. The nonlinear 3D ORA filter which mention from the last journal it state that it is one of the very good filter applied on clinical study. To achieve of reducing tube voltage, filters can provide the good tool to enhance the resolution contrast information for clinician or radiologist to provide better diagnosis with the digital images through the display monitor. Other than that by means of the advanced processors and software algorithm, we can easily to apply difference filters on the digital images through the computers to find out the properties of difference filters.

From this experimental task using Matlab, which is a mathematics tool to apply on the original image. By comparing with mean square error (MSE analysis) by subtraction with original & filtered image together, one linear filter (average filter) and one nonlinear filter (median filter) by selecting the difference block size evaluate which optimum filter block size, which will bring out the best MSE (lowest) value. Base on the definition of the Mean Square Error – the coefficient represent how much difference between the filtered image compare with the original real image.
Results and Discussion

All filters can suppress some of the signal noise on the abdomen CT image. Assessment of noise suppression and similarity between original and filtered images interpret by Mean Square Error analysis on MatLab program. By this means, we can find out which is the suitable block size of the median filter which can greatly improved the image quality proved by using subtraction of original image and filtered image with their MSE value.

The below figure 1 is the comparison table between original image with filtered images. The figure 2 is the comparison table between subtraction images between original images with filtered images.

Figure 1

Figure 2

Linear Filter

In general, the use of linear filter can retain image contrast. For averaging filter, loss of fine details appearance poorer resolution. Edge detection is also poorer.

Nonlinear Filter

In median filter, contrast of the CT image can be retained. Still, loss of the fine details, Improve the spatial resolution appearance “Mean Square Error Value” by select the optimum filter block size. The detail we can refer to the below Table 1.

Median Filter Block size	2x2	3x3	4x4	5x5	6x6	7x7	8x8	Ave Filter
Mean Square Valu e	12.24	6.60	14.72	12.95	17.54	16.49	19.42	10.78

Table 1

From the above, it is noted that in using noise reduction filter, no matter linear or nonlinear, a concerned with the resultant image quality should be improved. The above result can have an implication on use of different noise-reducing filters in different circumstances. Overall, the nonlinear median filter (3x3) has the best performance according to the lowest MSE value. To sum up, base on the characteristic of different filters, apply the linear filter (average filter) for the CT abdomen image on Pathology diagnosis, because it can let the image signal more homogeneity then better to appear the small spatial object such as malignant tumor seed. On the contrary, we can apply the nonlinear filter (median filter) for the CT abdomen image on anatomy diagnosis, because this filter can enhance the spatial contrast better for anatomy contour enhancement.