Adaptive morphology structural element construction algorithm based on local pixel density and symmetry

doi:10.1007/s11042-022-13259-3

2022-06-04

PRE

房超 (Chao Fang)

王小鹏 (Xiaopeng Wang) *

Qingsheng Wang

DOI：10.1007/s11042-022-13259-3

原文链接

原文求助

摘要

En 中文

Conventional morphological operations are implemented with fixed shape and size structuring elements. However, under these conditions, image features may change due to the complexity and diversity of images, leading to target boundary shifting, pseudo-target appearance, loss of detail, etc. In this paper, an adaptive mathematical morphology structuring element based on local pixel density and symmetry is proposed. The proposed structuring element can adaptively change shape according to the local pixel density, symmetry, and boundary features of pixels of image targets. First, the image is locally smoothed with a neighbourhood grey difference changing vector field to produce homogeneous image targets. Second, a local pixel density function based on pixel Euclidean distance, a pixel symmetry function based on pixel relative density, and a variation coefficient based on local pixel density and symmetry are defined. Finally, whether the centre pixel of the local pixel neighbourhood is a border pixel is determined with the variation coefficient. The adaptive structuring element is composed of all the border pixels determined in the local region of the image. According to the constructed structuring element, morphological dilation and erosion as well as other derived morphological operations are presented. The experimental results show that, with morphological operations, the proposed structuring element has the ability to adaptively fit the shape of image targets, preserve the necessary information of image features completely, and avoid image target boundary shifting. In addition to weakening the shading details, other image information is preserved, and distortion of the image is reduced.

Keyword:

Mathematical morphology

Adaptive structuring element

Local pixel density

Symmetry