Object-Centered Image Stitching

Herrmann, Charles; Wang, Chen; Bowen, Richard Strong; Keyder, Emil; Zabih, Ramin

doi:10.1007/978-3-030-01219-9_50

Cited by 26 publications

(24 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Currently, the global-local weight w (see Equation ( 4) is still decided manually by the user. Object recognition [33] or line detection [34] to find the focused object region can automatically adjust global-local weight w. We can group the vertices according to objects detected in the region. We then use the same local model term for vertices that belong to the same objects.…”

Section: Qualitative Measurementmentioning

confidence: 99%

Planar-Equirectangular Image Stitching

2021

View full text Add to dashboard Cite

The 360° cameras have served as a convenient tool for people to record their special moments or everyday lives. The supported panoramic view allowed for an immersive experience with a virtual reality (VR) headset, thus adding viewer enjoyment. Nevertheless, they cannot deliver the best angular resolution images that a perspective camera may support. We put forward a solution by placing the perspective camera planar image onto the pertinent 360° camera equirectangular image region of interest (ROI) through planar-equirectangular image stitching. The proposed method includes (1) tangent image-based stitching pipeline to solve the equirectangular image spherical distortion, (2) feature matching scheme to increase correct feature match count, (3) ROI detection to find the relevant ROI on the equirectangular image, and (4) human visual system (HVS)-based image alignment to tackle the parallax error. The qualitative and quantitative experiments showed improvement of the proposed planar-equirectangular image stitching over existing approaches on a collected dataset: (1) less distortion on the stitching result, (2) 29.0% increased on correct matches, (3) 5.72° ROI position error from the ground truth and (4) lower aggregated alignment-distortion error over existing alignment approaches. We discuss possible improvement points and future research directions.

show abstract

Section: Qualitative Measurementmentioning

confidence: 99%

Planar-Equirectangular Image Stitching

2021

View full text Add to dashboard Cite

show abstract

“…An energy map was modified based on the human eye's perception defined by saliency and color differ-ence [16]. As advanced object detection techniques developed [30], [31], Herrmann et al composited the energy using the detected object information [17]. Table 1 summarizes the approaches to find seams and pros/cons of previous seam finding algorithms.…”

Section: A Image and Video Stitchingmentioning

confidence: 99%

“…It, however, limits many potential applications and usages that apply video stitching only to images with small parallax, lens distortion, scene motion, and exposure difference. Thus, many image composition algorithms, such as seam cutting [13][14][15][16][17], and advanced image blending [4], [18] algorithms, have been proposed to relieve this registration artifacts and produce visually plausible stitched images. In addition to the challenges of image stitching, video stitching further suffers from visual artifacts caused by temporal inconsistency and time constraints.…”

Section: Introductionmentioning

confidence: 99%

Segmentation-Based Seam Cutting for High-Resolution 360-Degree Video Stitching

et al. 2021

View full text Add to dashboard Cite

We present a novel segmentation-based seam cutting algorithm to generate visually plausible high-resolution 360-degree video efficiently. While the demand for an efficient video stitching algorithm for generating immersive videos has increased, it has received limited attention in the literature. Furthermore, stitched videos often suffer from distorted objects, temporal inconsistency, and time constraints. Thus, in this paper, we propose an efficient seam finding algorithm that preserves objects from distortion, minimizes temporal inconsistency and reduces processing time. One of the fundamental steps in image and video stitching is the estimation of seam boundary. To do this, the proposed algorithm leverages a convolutional neural networks-based instance segmentation algorithm that provides more accurate object regions. It computes energy surfaces considering the regions and then estimates seam boundary by discovering a minimal energy path with minimal computations. We validate the proposed algorithm using challenging high-resolution 360-degree virtual reality sequences. The experimental results verify that the proposed algorithm can produce seam boundaries that avoid objects with better temporary consistency. The proposed algorithm reduces the number of pixels passed through objects by approximately 30% on average compared to the existing algorithms. The qualitative comparisons furthermore demonstrate that the proposed algorithm consistently produces more perceptually pleasing results.

show abstract

“…The line matching method has obvious advantages, but the high computational complexity limits its application range. Herrmann et al [ 31 ] made full use of object detection [ 32 ] and combined the multiple registration algorithm [ 11 ] to construct an object-centric image mosaic framework. Multiple potential planes generated by multiple registration can effectively solve the occlusion of foreground objects on the background, but it also makes the search of seam lines more complicated.…”

Section: Related Workmentioning

confidence: 99%

Image Stitching Based on Nonrigid Warping for Urban Scene

Deng

Yuan

Deng

et al. 2020

Sensors

View full text Add to dashboard Cite

Image stitching based on a global alignment model is widely used in computer vision. However, the resulting stitched image may look blurry or ghosted due to parallax. To solve this problem, we propose a parallax-tolerant image stitching method based on nonrigid warping in this paper. Given a group of putative feature correspondences between overlapping images, we first use a semiparametric function fitting, which introduces a motion coherence constraint to remove outliers. Then, the input images are warped according to a nonrigid warp model based on Gaussian radial basis functions. The nonrigid warping is a kind of elastic deformation that is flexible and smooth enough to eliminate moderate parallax errors. This leads to high-precision alignment in the overlapped region. For the nonoverlapping region, we use a rigid similarity model to reduce distortion. Through effective transition, the nonrigid warping of the overlapped region and the rigid warping of the nonoverlapping region can be used jointly. Our method can obtain more accurate local alignment while maintaining the overall shape of the image. Experimental results on several challenging data sets for urban scene show that the proposed approach is better than state-of-the-art approaches in both qualitative and quantitative indicators.

show abstract

Object-Centered Image Stitching

Cited by 26 publications

References 28 publications

Planar-Equirectangular Image Stitching

Planar-Equirectangular Image Stitching

Segmentation-Based Seam Cutting for High-Resolution 360-Degree Video Stitching

Image Stitching Based on Nonrigid Warping for Urban Scene

Contact Info

Product

Resources

About