Machine Vision and Applications (2021) 32:5
https://doi.org/10.1007/s00138-020-01130-0
ORIGINAL PAPER
A novel approach for unsupervised image segmentation fusion
of plant leaves based on G-mutual information
Navid Nikbakhsh
1
· Yasser Baleghi
1
· Hamzeh Agahi
2
Received: 14 November 2019 / Revised: 2 June 2020 / Accepted: 15 September 2020
© Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract
Plant leaf segmentation has a very important role in most plant identification methods. Tree leaves segmentation in images
with complex background is very difficult when there is no prior information about the leaves and backgrounds. In practice,
the parameters of unsupervised image segmentation algorithms must be set for each image to get the best results. In this
paper, to overcome this problem, fusion of the results of five leaf segmentation algorithms (fuzzy c-means, SOM and k-
means in various color spaces or different parameters) is applied. To fuse the results of these segmentations, new equations
for mutual information (g-mutual information equations) based on the g-calculus are introduced to find the best consensus
segmentation. The results of the mentioned primary clustering algorithms are considered as a new feature vector for each
pixel. To reduce the time complexity, a fast method is employed using truth table containing different feature vectors. To
evaluate this new approach, a leaf image database with natural scenes, taken from Pl@ntLeaves database, is generated to have
different positions and orientations. In addition, a widely used database is used to compare the proposed method with other
methods. The experimental results presented in this paper show that the use of g-calculus in fusion of image segmentations
improves the evaluation parameters.
Keywords Pseudo-operations · Image segmentation fusion · G-calculus · Mutual information · Tree leaves segmentation
1 Introduction
Plant classification is the foundation of plant ecology, plant
medicine, plant genetics and life science. It plays an impor-
tant role in the protection, development and exploitation of
plant resources. Plant classification is not only tedious and
time consuming, but also requires special expertise. Image
processing and machine vision are electronic engineering
techniques that can provide promising tools for automatic
plant identification. Machine learning allows the machine
to learn from examples and experience. The general goal
of machine learning and machine vision is to automate the
extraction of information from images, in order to reach
high-level understanding from the content of the images
B Yasser Baleghi
y.baleghi@nit.ac.ir
1
Department of Electrical and Computer Engineering, Babol
Noshirvani University of Technology, Shariati Ave, Babol
47148-1167, Iran
2
Department of Mathematics, Faculty of Basic Science, Babol
Noshirvani University of Technology, Shariati Ave, Babol
47148-71167, Iran
and perform tasks such as classification and identification
of images content.
Efficient plant segmentation is one of the most critical
steps that can help to accurately label images for detection
and control of plants, weeds and fruits [1, 2]. Leaves are
the most important part of plants for identification, due to
their proper characteristics such as access throughout the year
and the simplicity of their analysis through two-dimensional
images. In most approaches, accurate extraction of the shape
of a leaf in images with natural background is very important
for classification, recognition of plant diseases, and mon-
itoring plant growth. Today, due to the development and
availability of mobile devices and remote access to databases,
there is a growing interest in automating the process of
plant identification from images taken in natural area. Many
databases have tree leaf images obtained under controlled
conditions like uniform light and white background. In these
images, leaf extraction using methods such as the Otsu tech-
nique [3] or fixed thresholds is very easy and shows good
accuracy. However, the automatic segmentation of the tree
leaves from images with natural background and uncon-
trolled imaging conditions is a very challenging task, as
0123456789().: V,-vol 123