Sharpening methods for low-contrast images based on nonlocal differences

Abstract

Various phenomena, such as quantum noise and scattering,

existing in the industrial X-ray imaging process, and structural

complexity of the measured workpiece resulted in low-contrast

and blurry industrial X-ray images, which caused interference to

the X-ray image analysis. This study proposed an improved adap

tive sharpening algorithm to enhance the contrast and quality of

X-ray images. Image pixels and the neighborhood established a

non-local feature relationship through non-local filtering model

on the basis of the structural similarity index measure (SSIM)

model. The structural similarity of block-based pixels within the

search window area was calculated. The improved weight was

combined with image sharpening with high-enhancement re

sults. The edge-preserving ability of the algorithm was verified

through image tests. Finally, the proposed algorithm contributed

to improving the quality of the contrast of industrial X-ray im

ages through simulation experiments. Results demonstrate that

features in the neighborhood based on non-local differences re

flect rich details of images. The X-ray images sharpened with the

proposed algorithm are characterized with excellent visual effects

and rich details, with information entropy (IE) values of 2.1464,

4.2453, and 3.7283 and structural similarities of 0.9521, 0.9238,

and 0.9534. The weights calculated by SSIM indicate that a high

similarity in structure exists between the sharpened image and

the original one. Images processed by the sharpening algorithm

based on non-local differences present prominent details while

effectively maintaining the objective parameter values. This study

provides references to improve the quality of low-contrast images.