An analysis of a un decimated BM3D de-speckling method for monochrome images

Author Name(s): C.Chandra Sekhar, Y.Murali Mohan Babu, Mahesh
Author Email: umashekar_2000@yahoo.com

Abstract

In microwave or synthetic aperture radar or radar image formation, a speckle noise is generally added. This noise is available in coherent imaging systems like radar image capturing, medical image capturing, etc. De-speckling methods are used to de-speckle the available speckle noise from these speckled systems. In optical image formation, other noises are added.An un-decimated block matching 3D (BM3D) method is a state of art technique in de-speckling area. In this paper, an analysis has been done on optical or satellite images with un-decimated BM3D method. Different quality parameters are evaluated and compared for different speckle noise variances by considering the Sentinel-2 satellite sub band images as radar image.

Introduction

Microwave images are popular in taking the images at bad environmental situations and at day and nights. In this radar image process of capturing or image formation, a noise is generally added called as speckle noise. Due to this unavoidable noise the quality of the image is degraded and the important features are disturbed. To get back the original image of the scene, a good de-speckling method [1] has to be used. In this process, spatial domain filters like lee, map and wiener filters and frequency domain filters like wavelet, ridgelet, curvelet, bandlet, contourlet transforms are generally used [2-3]. A BM3D technique has become popular in removing the speckle noise in last decade. An un-decimated BM3D algorithm is giving good and reasonable results in recent years. A de-speckled image can be used for applications like ground water monitoring, agriculture advising, disaster management, forest deformation, glacier analysis, etc. Here, a question arises. Can a de-speckling technique is used for an optical image, by considering the optical sub band as radar image. This paper gives answer for that.

Conclusion

The paper has been given the statistical approach to un-decimated BM3D transformation as de-speckling method for de-speckling of different mono chrome images. It has been considered the noise variance from 0.25 to 0.05 for speckling purpose. It is a trial and error process of using de-speckling method for optical images. The de-speckling method is generally meant for microwave images. It has been successfully executed and given the good results even for optical images, if they have been converted into mono chrome images. The mono chromatic images like red, green and blue bands and grey colour images have been tested for de-speckling. All the quality parameters like ENL, SSI, SMPI, CC, ESIH, ESIV, MSE, PSNR, EPI and CT have been calculated for four images of two sets. Generally ENL is calculated for microwave or radar images. This value is generally high for radar images.

Except EPI, all other results are good. The lesser value of EPI may be due to low resolution of the images taken. It is also an important feature that an image should have. The computational time may be more but it is not major factor to discuss about a de-speckling method. It is concluded that the given de-speckling is good to use for any mono chromatic images. But, this analysis has been done for less number of images, so to give the final conclusion it has to be tested for more images of more varieties.

References

[1]. IqbalM, Chen J, Yang W, Wang P,  Sun B. SAR image de-speckling by selective 3D filtering of multiple compressive reconstructed images. Progress In Electromagnetics Research, Vol. 134, No. 12, 209-226, 2013.

[2]. Murali Mohan Babu Y, Subramanyam MV,Giriprasad MN. A New Approach For SAR Image De-noising. International Journal of Electrical and Computer Engineering, Volume 5, Issue 5, 984-991, October 2015.

[3]. Stankiewicz A, Marciniak T, DąbrowskiA, Stopa M, Rakowicz P, Marciniak E. De-noising methods for improving automatic segmentation in OCT images of human eye. Bulletin of The Polish Academy of Sciences Technical Sciences, Vol. 65, No. 1, 71-78, 2017.

[4]. Subramanyam MV,MM Babu Y, Giriprasad MN. A Modified BM3D Algorithm For SAR Image De-speckling.Procedia Computer Science (Elsevier), Volume 70, Issue 1, 69 – 75, December 2015.

[5]. Qingsong Zhu, Jiaming Mai, Ling Shao. A Fast Single Image Haze Removal Algorithm Using Color Attenuation Prior. IEEE Transactions on Image Processing, Volume 24, Issue 11, 3522 – 3533, November 2015.

[6].Kai Zhang,  WangmengZuoYunjin Chen. Beyond a Gaussian De-noiser: Residual Learning of Deep CNN for Image De-noising.IEEE Transactions on Image Processing, Volume: 26, Issue: 7, 3142-3155, July 2017.

[7]. Gabriela Ghimpe¸teanu, Thomas Batard, Marcelo Bertalmío, Stacey Levine. A Decomposition Framework for Image De-noising Algorithms. IEEE Transactions on Image Processing, Volume. 25, No. 1, 388- 399, January 2016.

[8]. Murali Mohan Babu Y,  Radhika K. A new approach for microwave imagery de-noising. International Journal of Image, Graphics and Signal Processing, Volume 5, issue 1, 52-60, May 2016.

[9]. MitraFatemi, ArashAmini, LoicBaboulaz,  MartinVetterli. Shapes From Pixels.IEEE Transactions on Image Processing, Vol. 25, No. 3, 1193-1203, March 2016.

[10]. AbderrahimHalimi, Gerald Buller S, Stephen McLaughlin, Paul Honeine.De-noising Smooth Signals Using a Bayesian Approach: Application to Altimetry. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Volume 10, No. 4, 1278-1289, April 2017.

793 total views, no views today

Download PDF File

About the author: admin