Microscopy mineral image enhancement based on improved adaptive threshold in nonsubsampled shearlet transform domain

Liangliang Li,Zhenhong Jia,Yujuan Si

doi:10.1063/1.4998400

Liangliang Li, Zhenhong Jia + Show 1 more

Open Access

https://doi.org/10.1063/1.4998400

Copy DOI

Journal: AIP Advances	Publication Date: Mar 1, 2018
Citations: 8	License type: cc-by

Affiliation: Jilin University, Xinjiang University

Abstract

In this paper, a novel microscopy mineral image enhancement method based on adaptive threshold in non-subsampled shearlet transform (NSST) domain is proposed. First, the image is decomposed into one low-frequency sub-band and several high-frequency sub-bands. Second, the gamma correction is applied to process the low-frequency sub-band coefficients, and the improved adaptive threshold is adopted to suppress the noise of the high-frequency sub-bands coefficients. Third, the processed coefficients are reconstructed with the inverse NSST. Finally, the unsharp filter is used to enhance the details of the reconstructed image. Experimental results on various microscopy mineral images demonstrated that the proposed approach has a better enhancement effect in terms of objective metric and subjective metric.

Highlights

Microscopy imaging is an effective tactics to analysis the microstructure of minerals.[1,2] due to improper use of the microscopy or unexpected external effects in the collection of mineral samples, the resolution of the mineral images obtained under the microscope is not high.[3]
We will simulate on the mineral image database from http://webmineral.com/ to demonstrate the effectiveness of the proposed enhancement method based on non-subsampled shearlet transform (NSST)
The approaches, including the traditional histogram equalization method (HE),[27] contrast limited adaptive histogram equalization method (CLAHE),[28] the image enhancement algorithm based on nonsubsampled shearlet transform and guided filtering (NSST-GF),[20] the edge-based texture histogram equalization method (ETHE),[29] the dominant orientation-based texture histogram equalization method (DOTHE),[29] the feature-linking model for image enhancement (FLM),[30] the linking synaptic computation network for image enhancement (LSCN),[31] are applied as the comparison techniques

Summary

Introduction

Microscopy imaging is an effective tactics to analysis the microstructure of minerals.[1,2] due to improper use of the microscopy or unexpected external effects in the collection of mineral samples, the resolution of the mineral images obtained under the microscope is not high.[3]. Contourlet transform has been widely used in image fusion and enhancement, and has achieved good results.[11,12] due to the existence of up-sampling and down-sampling mechanism in LP and DFB, the contourlet does not have translation invariance. In order to solve this problem, non-subsampled contourlet transform (NSCT) is proposed,[13,14,15] which is based on the extension of contourlet. NSCT has translation invariance, it can overcome the Gibbs effect of previous transformations, but it has too many computational data and high computational complexity, so it is difficult to meet the requirements of real-time applications. Compared to NSCT, the enhancement method based on shearlet transform (ST) has more flexible structure, higher computational efficiency and better image enhancement effect, but it does not have translation invariance.[16]

Methods

Results

Conclusion