A Wavelet Transform-Based Image Mosaic Algorithm
From the basis of studying in SIFT (Scale-Invariant Feature Transform) operator and wavelet transform, a multi-scale image mosaic algorithm is proposed. The proposed mosaic algorithm maintains the advantage that SIFT operator is invariant to scaling and r
- PDF / 222,805 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 10 Downloads / 280 Views
Abstract From the basis of studying in SIFT (Scale-Invariant Feature Transform) operator and wavelet transform, a multi-scale image mosaic algorithm is proposed. The proposed mosaic algorithm maintains the advantage that SIFT operator is invariant to scaling and rotation transformations, combining with the wavelet transform image fusion method, it enhance the anti-noise ability, maintains a certain stability to brightness change, improved the joint effect. Keywords SIFT operator
⋅
Image mosaic
⋅
Wavelet transform
1 Introduction Image mosaic is a integral image technology, through a series of operations include space registration, image transform, resampling, and image fusion, split two or more image which have overlapping portions into a large seamless image [1]. Image mosaic has a very wide range of applications, such as computer vision, resolution enhancement, medical imaging, and so on. Commonly, mosaic involves 3 steps: image preprocessing, image registration, image fusion. Image registration is one of the most important and foundation step which use to estimate the geometric transformation relationship of a set of images. Then, reproject the image in a big background merging both the overlapping portions and no overlap ones. The next step is to the discontinuities in the boundary between two images.
Q. Sun (✉) ⋅ H. Yang Department of Electronic Engineering, Anhui Technical College of Mechanical and Electrical Engineering, Wuhu, Anhui, China e-mail: [email protected] Q. Sun School of Automation and Electronic Engineering, Tianjin University of Technology and Education, Tianjin, China © Springer Nature Singapore Pte Ltd. 2018 Q. Liang et al. (eds.), Communications, Signal Processing, and Systems, Lecture Notes in Electrical Engineering 423, https://doi.org/10.1007/978-981-10-3229-5_95
881
882
Q. Sun and H. Yang
In recent years, many researchers proposed a lot of methods in image registration and fusion. Compared with many traditional registration algorithms, feature-based method such as SIFT and SURF detectors proved to be more efficient, more accurate, more reliable registration algorithm. This paper combines wavelet transform and SIFT operator, proposed an image mosaic algorithm.
2 Image Registration Based on SIFT Image registration is the align process in space of over laying two or more images of the same scene taken at different times, from different view points, or by different sensors [1]. Image registration based on SIFT first finds the corresponding feature points of the overlapping parts of the two images, then calculated the transformation relation between the images by using these features, and then transformed the two images into the same coordinate system. SIFT operator, called the scale-invariant feature transform, proposed by Lowe [2, 3], which is an image local feature descriptor based on the scale space theory. SIFT feature points on the image have invariance to scale and rotation, scaling, strong noise resistance characteristics. 1. Feature detection Using Gauss’s function as the convol
Data Loading...
