Surface abnormality detection in medical and inspection systems using energy variations in co-occurrence matrixes

TL;DR

This paper introduces a surface defect detection method using energy variations in co-occurrence matrices, achieving high accuracy and low computational cost, applicable in industrial and medical imaging.

Contribution

The novel approach utilizes energy changes in co-occurrence matrices for defect detection, with a training-testing framework adaptable to various image types.

Findings

Achieved approximately 89.56% accuracy on 2d-hela dataset.

Demonstrated high detection accuracy on stone and ceramic images.

Method offers low computational complexity and broad applicability.

Abstract

Detection of surface defects is one of the most important issues in the field of image processing and machine vision. In this article, a method for detecting surface defects based on energy changes in co-occurrence matrices is presented. The presented method consists of two stages of training and testing. In the training phase, the co-occurrence matrix operator is first applied on healthy images and then the amount of output energy is calculated. In the following, according to the changes in the amount of energy, a suitable feature vector is defined, and with the help of it, a suitable threshold for the health of the images is obtained. Then, in the test phase, with the help of the calculated quorum, the defective parts are distinguished from the healthy ones. In the results section, the mentioned method has been applied on stone and ceramic images and its detection accuracy has been calculated and compared with some previous methods. Among the advantages of the presented method, we can mention high accuracy, low calculations and compatibility with all types of levels due to the use of the training stage. The proposed approach can be used in medical applications to detect abnormalities such as diseases. So, the performance is evaluated on 2d-hela dataset to classify cell phenotypes. The proposed approach provides about 89.56 percent accuracy on 2d-hela.