| Pub. Date | : Oct, 2019 |
|---|---|
| Product Name | : The IUP Journal of Electrical and Electronics Engineering |
| Product Type | : Article |
| Product Code | : IJEEE11910 |
| Author Name | : Sandip R Panchal, Jaymin K Bhalani and Vrushank H Gandhi |
| Availability | : YES |
| Subject/Domain | : Engineering |
| Download Format | : PDF Format |
| No. of Pages | : 13 |
Image registration is a fundamental image processing task to match and align physically two images which could have been imaged by different sensors, view angles and/or at different times. While registering the satellite images, there are several unique challenges like cloud pixels, noise in the images, systematic errors, multispectral images, terrain induced distortions, etc. Through image registration techniques, we are finding a proper geometric transform between two images that can align corresponding points in them. It is the foundation of applications, such as image fusion, medical image processing, remote sensing and threedimensional (3D) image reconstructions. Remote sensing is one of the fields that have benefited most from image registration technique for stitching the reference and sensed images to get a larger view of the scene. The paper presents the Least Square-Support Vector Machine (LS-SVM) which has very strong theoretical foundation based on statistical learning theory. The experimental results demonstrate that global deformations can be well registered by LS-SVM-based multiview image registration.
Multiview image registration, also referred to as image mosaicing (stitching), is mostly used in remote sensing field where merging of two or more satellite images is very frequently required (Zitova and Flusser, 2003). In addition to this, if the algorithm is fully automatic and can tackle the different global geometric distortions which might occur very frequently in multiview image registration, then it will always be advantageous for us. The possible global nonlinearities are affine deformations (like scaling, clockwise and counter clockwise rotations, different degrees of horizontal and vertical shears) and perspective projections and so they must be tackling without human interventions (Medha et al., 2009).
Image registration, Image stitching, Machine learning, Least Square-Support Vector Machine (LS-SVM), Support Vector Machine (SVM)