This paper presents a unified method based on pixels for identifying the geometric deformations of digital images. This method uses radiometric pixel gray values to represent the geometric deformations over the entire image surface, a test plate with a 38 × 22 grid, along x and y directions. This method uses a test plate to measure the radiometric pixel gray values in images which represent the geometric deformations in the image. For image registration and the detection of the geometric deformations the following six geometric transformation methods were utilized; non-reflective similarity, similarity, affine, projective, polynomial, piecewise linear and three resampling methods; nearest neighbour, bilinear and bicubic were used. The image data were taken by using Olympus E-150 digital SLR camera and the applied geometric transformation and resampling methods were coded in Matlab software. The experimental results of all the methods are presented and evaluated. The results showed that non-reflective similarity, similarity and affine transformations have a better accuracy than the other methods. Furthermore, geometric distortions were calculated by using corresponding grid corners of pixel coordinates in normal and registered images. Among all geometric transformation, projective transformation combined with three resampling methods revealed the best results.
Key words: Digital image, transformation, geometric deformations, registration, resampling.
Copyright © 2022 Author(s) retain the copyright of this article.
This article is published under the terms of the Creative Commons Attribution License 4.0