Local-Adaptive Image Alignment Based on Triangular Facet Approximation

Locally Aligned Image Stitching Based on Multi-Feature and Super-Pixel Segmentation With Plane Protection

“…This section describes some experiments performed to evaluate the performance of the method on a series of challenging images. We compare our approach with the classical local homography methods [13,16,18] and other state-of-the-art stitching algorithms [27,28,33]. In our experiments, we used SIFT [35] and LSD [19] to detect the corresponding point and line features and matched the SIFT features through GMS [39], and HLM [42] matched the LSD features.…”

“…In the previous work, we optimized the feature fine matching, and inspired by [27], we proposed the As-aligned-as-possible (AAAP) warp [28] based on the thin-plate spline (TPS) [29] to correct the mapping deviation to obtain more accurate stitching of images. In the latest work of Li et al [33], the scene is approximated as a combination of adjacent triangular facets, and a warping based on the triangular facet approximation (TFA) is proposed.…”

“…Moreover, the optimized warping for the homography model [9], [31] effectively solves the distortion problem caused by the change of viewing angle. Moreover, reasonable meshing to adapt to the coplanar characteristics of homography can also achieve better visual effects [32]- [33]. The warps have achieved considerable splicing results, but they are difficult to cope with complex scenes (see Figure 1(e)).…”

Performance Evaluation of Bundle Adjustment with Population Based Optimization Algorithms Applied to Panoramic Image Stitching

“…For the local exploitation stage, a new bat is generated according to a solution selected among the current best solutions, using a random walk as described in Equation (13).…”

“…Some questions remain unresolved and from a computational geometry perspective the result usually contains cut objects or blur [ 12 ]. The image stitching method is the joining of images based on their overlapping areas and can be defined in two main steps: alignment and blending [ 13 ].…”