Semantic segmentation guided feature point classification and seam fusion for image stitching

Image stitching can be employed to stitch images taken from different times, perspectives, or devices into a panorama with a wider view. However, the imaging specification of images to be stitched is strict. If the imaging specification is not satisfied, artefacts caused by inaccurate alignment and unnatural distortion will occur. Semantic segmentation can solve the classification problem at the pixel level; however, image stitching significantly depends on the accuracy of feature points. Therefore, this paper proposes an image stitching algorithm based on semantic segmentation to guide feature point classification and seam fusion. First, the images are recognized by a cascade semantic segmentation network, and the image feature points are classified. Thereafter, the corresponding homography transformations are calculated using different class feature points, and the best homography mapping for the entire target image is selected. Finally, a seam-cutting algorithm based on semantic segmentation is used to compute the seam, and a feathering Poisson fusion with distance transformation is used to eliminate artefacts and light differences. Experiments show that the algorithm can generate transitional natural and perceptual stitching results even under the influence of perspective and light differences.

Quasi Cubic Trigonometric Curve and Surface

Firstly, a new set of Quasi-Cubic Trigonometric Bernstein basis with two tension shape parameters is constructed, and we prove that it is an optimal normalized totally basis in the framework of Quasi Extended Chebyshev space. And the Quasi-Cubic Trigonometric Bézier curve is generated by the basis function and the cutting algorithm of the curve are given, the shape features (cusp, inflection point, loop and convexity) of the Quasi-Cubic Trigonometric Bézier curve are analyzed by using envelope theory and topological mapping; Next we construct the non-uniform Quasi-Cubic Trigonometric B-spline basis by assuming the linear combination of the optimal normalized totally positive basis have partition of unity and continuity, and its expression is obtained. And the non-uniform B-spline basis is proved to have totally positive and high-order continuity. Finally, the non-uniform Quasi Cubic Trigonometric B-spline curve and surface are defined, the shape features of the non-uniform Quasi-Cubic Trigonometric B-spline curve are discussed, and the curve and surface are proved to be continuous.

Three-dimensional garment pattern design using progressive mesh cutting algorithm

Purpose The purpose of this paper is to develop the core module of computer-aided three-dimensional garment pattern design system. Design/methodology/approach A progressive mesh cutting algorithm and mesh reshaping algorithm have been developed to cut a single mesh into multiple patches. A flat projection algorithm has been developed to project 3D patches into 2D patterns. Findings The software developed in this study is expected to enable its users to design complex garment patterns without the in-depth knowledge of pattern design process. Research limitations/implications The mesh model used in this study was a fixed model. It will be extended to a deformable garment model that can be resized according to the underlying body model Practical implications The software developed in this study is expected to reduce the time required for time-consuming and trial-and-error-based pattern design process. Social implications Fashion designers will be able to design complex patterns by themselves and the dependence upon expert patterners could be reduced Originality/value The progressive mesh cutting algorithm developed in this study can cut a mesh model using arbitrary lines. The mesh reshaping algorithm can improve the mesh quality of divided patches to increase the numerical stability during subsequent pattern flattening process. The flip removal algorithm can effectively remove the partially flipped mesh elements.

A Class of Trigonometric Bernstein-Type Basis Functions with Four Shape Parameters

In this work, a family of four new trigonometric Bernstein-type basis functions with four shape parameters is constructed, which form a normalized basis with optimal total positivity. Based on the new basis functions, a kind of trigonometric Bézier-type curves with four shape parameters, analogous to the cubic Bézier curves, is constructed. With appropriate choices of control points and shape parameters, the resulting trigonometric Bézier-type curves can represent exactly any arc of an ellipse or parabola. The four shape parameters have tension control roles on adjusting the shape of resulting curves. Moreover, a new corner cutting algorithm is also proposed for calculating the trigonometric Bézier-type curves stably and efficiently.

New Trigonometric Basis Possessing Denominator Shape Parameters

Four new trigonometric Bernstein-like bases with two denominator shape parameters (DTB-like basis) are constructed, based on which a kind of trigonometric Bézier-like curve with two denominator shape parameters (DTB-like curves) that are analogous to the cubic Bézier curves is proposed. The corner cutting algorithm for computing the DTB-like curves is given. Any arc of an ellipse or a parabola can be exactly represented by using the DTB-like curves. A new class of trigonometric B-spline-like basis function with two local denominator shape parameters (DT B-spline-like basis) is constructed according to the proposed DTB-like basis. The totally positive property of the DT B-spline-like basis is supported. For different shape parameter values, the associated trigonometric B-spline-like curves with two denominator shape parameters (DT B-spline-like curves) can be C2 continuous for a non-uniform knot vector. For a special value, the generated curves can be C(2n-1) (n=1,2,3,…) continuous for a uniform knot vector. A kind of trigonometric B-spline-like surfaces with four denominator shape parameters (DT B-spline-like surface) is shown by using the tensor product method, and the associated DT B-spline-like surfaces can be C2 continuous for a nonuniform knot vector. When given a special value, the related surfaces can be C(2n-1) (n=1,2,3,…) continuous for a uniform knot vector. A new class of trigonometric Bernstein–Bézier-like basis function with three denominator shape parameters (DT BB-like basis) over a triangular domain is also constructed. A de Casteljau-type algorithm is developed for computing the associated trigonometric Bernstein–Bézier-like patch with three denominator shape parameters (DT BB-like patch). The condition for G1 continuous jointing two DT BB-like patches over the triangular domain is deduced.