Measurement of image quality is very important to manyimage processing systems. Since it inherent physical limitations and economicreasons for the quality of images and videos, such that it could view by ahuman observer.
Some of existing measures of image quality are listed below.Imagedifferencing is used to determine changes between images and Maximumdifference (MD) is themaximum of the error between real image and filtered image. For the betterperformance maximum difference should be minimum and large value shows poorquality.
The root mean square error (RMSE) is the square root of cumulativesquared error between the filtered and the original image for comparing variousimage processing. For thebetter performance root mean squareerror should be minimum and large value shows poor qualityThe peak signal-to-noise ratio(PSNR) inimage processing is a standard measure of the sensitivity ofimaging system. For the betterperformance peak signal-to-noiseratio should be maximum and smaller value shows poor quality.In image processing normalized absolute error (NAE)is a measure of difference between thefiltered and the original image for comparing various images. Higher value ofnormalized absolute error is used for poor quality of images.Smaller the value of structural content (SC) better is theimage quality and higher value of structural content is used for poor qualityof images.
For image-processingapplications in which the brightness of the image and template can vary due tolighting and exposure conditions, the images can be first normalized. Normalized Cross Correlation (NCC) showbetter quality with higher value.Laplacianmean squared error (LMSE) is based on the importance of edgesmeasurement with lower the value of laplacian mean squared error better thequality of images.Universalimage quality index (IQI) is a image quality distortion measurement which is definedas the product of three factors: structural distortion, contrast distortion andluminance distortion, of the distortion of ideal image with respect to filteredimage.The Structural Similarity (SSIM) index is a method for measuresimilarity between ideal image and filtered image. It is an quality measure ofone of the filtered images being compared, providedthe ideal image is regarded as of perfect quality . SSIM is a better than thatof universal image quality index. The Pratt’s Figure of Merit(PRATT) find the edge location accuracy bythe displacement of filtered image from an ideal image where both the image(filtered & ideal) are edge detected images from the edge detectors likesobel , prewitt , Robert , log etc.
.The Mammographic Image Analysis Society (MIAS)database is the largest publicly available database of mammographic data. Itcontains approximately 322 screening mammography cases, where 207 imagesrepresent normal, while 64 and 51 images referred as benign and malignant casesrespectively.For experimental purpose, the 45 images are taken fromMIAS database which includes of 15 normal images and 30 abnormal images. Theabnormal images are again classified into two classes which are benign andmalign. There are 15 benign images and 15 malign images.
In this section, the results of those filters are compared which arediscussed in previous section. Many proposed work in the literature are alsodiscussed and compared with the present approach. Furthermore, the filteredimages shown in this section are obtained as a result of subjecting the breastmammogram shown in Figure1 to variousfilters.The image quality metrics of median filterare presented in Table 1 and the results of median filter applied to theoriginal mammogram of Figure1 are shown in Figure 2 .
From the image quality metrics table it isfound that, the image quality degrades with the increase in windows size. Someof the performance of image metrics (MD,NAE,LMSE,RMSE,SC) increase and some(PSNR,IQI, SSMI,NCC) decreases with the increase of windows size .Also prattsfigure of merit show better performance with lower window size.Table2 shows the image quality metrics obtained for Adaptivemedian filter (AMF) and Figure 3 shows the results for AMF. In AMF the imagequality will not change much with the increase in windows size .
It is foundthat performance of AMF is quite good resulting in lower value of RMSE andhigher value of PSNR. The Pratts figure of merit also shows better results as compared tomedian filter.Figure 4 shows the result of frost filterand the image quality metrics obtained for frost filter is presented in Table 3. Here theimage quality will change variably with the increase in windows size .It isfound that performance of frost filter is quite good resulting in window size5.
Figure 5 shows the filtered images from waveletfiltering, and the corresponding tables of image quality metrics are presentedin Table 4. It is found that image quality ismaintained after filtering at first-level decomposition as indicated by RMSE,PSNR, IQI and SSMI while after second-level decomposition, image becomes muchbrighter and Pratt’s figure of merit reduces. When LH band is eliminated aftersecond-level decomposition, most of the details are lost giving MD of 0.000 andNAE of 254.
00 .The best result is obtained when HH band is eliminated afterfirst-level decomposition.The image quality metricsobtained for histogram equalization(column 2) & CLAHE(column 3) is presentedin Table 5 and the results for histogramequalization (a)& CLAHE (b)are shown in figure 6. It is found that withmaximum difference and psnr in histogram equalization is smaller than all otherfilter.
In this paper, review and comparison of representative denoisingmethods both qualitatively and quantitatively with extensive experimentsconduct to evaluate the performance of all the algorithms. In analyticalcomparison, it was found that image representations with over complete basisfunctions improve the performance within each category.In this paper it is clear from the comparison that all thedenoising techniques are important for various applications. In applicationsthat require high efficiency, some filters are used, some filters are moreappropriate for high searching complexity, memory and complexity issue .