Quantitative Dual Nature Analysis of Mean Square Error in SAR Image Despeckling | Use of MSE Metric in Case of With or Without Reference Index | Research Paper Discussion

Quantitative Dual Nature Analysis of Mean Square Error (MSE) in SAR Image Despeckling | Use of MSE Metric in Case of With or Without Reference Index. The MSE metric is one of the most common metrics being used in the field of image denoising. In the field of SAR image despeckling, this metric can be used in both the cases of with and without reference index. This is a very unique and useful study in the field of SAR image despeckling. The analysis is explained in the research paper being shown in Fig 1. This post will elaborate on those results of the research paper that can help the new researchers who are doing research in this field. If you have an interest in SAR image despeckling then stay here and read the content, it will be helpful to you. Let's begin...
Fig 1. Research paper by Dr. Prabhishek Singh and Dr. Raj Shree

The MSE metric is usually used as a quantitative metric in image denoising when the reference image is available. Then the MSE value of the denoised image is compared with the reference image. The comparative MSE value difference should be as minimum as possible. This shows better denoising results. But in the SAR image despeckling the MSE metric can also be used when the reference image is not available. How? Let's do the case study of some research papers.
In this case, when the reference SAR image is considered as speckle free. Now the despeckled SAR image is compared with this reference SAR image. Since PSNR is inversely proportional to MSE, therefore PSNR of the despeckled SAR image should be highest. This automatically means that the MSE of the despeckled SAR image should be minimum for the best despeckling results. The low MSE value denotes that there is very little difference between reference SAR image and despeckled SAR image. And small difference means better despeckling results.

So, in this case, the minimum value of the MSE metric depicts best despeckling results.

In this case, when a real speckled SAR image is available, and the despeckled SAR image is compared with the real speckled SAR image. In this case, also, the MSE metric can be used. The calculated MSE of despeckled SAR image wrt real speckled SAR image should be comparatively large. The comparative large difference in MSE value denotes better speckle reduction. The comparative difference should not be very large otherwise the despeckling results will degrade. 

The dual nature analysis of MSE metric in SAR image despeckling is analyzed in the below Table 1 at the noise variance of 20% using CASE 1 and CASE 2. study.


Fig 2. Research paper: Singh, P., Shree, R. A new homomorphic and method noise thresholding based despeckling of SAR image using anisotropic diffusion. Journal of King Saud University – Computer and Information Sciences (2017), http://dx.doi.org/10.1016/j.jksuci.2017.06.006


Fig 3. Research paper: Ashkan Masoomi, Roozbeh Hamzehyan, and Najmeh Cheraghi Shirazi, Member, IACSIT, “Speckle Reduction Approach for SAR Image in Satellite Communication”, International Journal of Machine Learning and Computing, Vol. 2, No. 1, February 2012.

The above study validates the dual nature of the MSE metric. In Fig 2, the best results are obtained at minimum MSE values in the case of speckle free reference SAR image. In Fig 3, the best results are obtained at maximum MSE values in the case of a real speckled SAR image.



                                    For more topics refer: https://www.gofastresearch.com/
Quantitative Dual Nature Analysis of Mean Square Error in SAR Image Despeckling | Use of MSE Metric in Case of With or Without Reference Index | Research Paper Discussion Quantitative Dual Nature Analysis of Mean Square Error in SAR Image Despeckling | Use of MSE Metric in Case of With or Without Reference Index | Research Paper Discussion Reviewed by Dr. Prabhishek Singh on June 08, 2020 Rating: 5

No comments:

Powered by Blogger.