MegaStitch: Robust Large-Scale Image Stitching

Abstract

We address fast image stitching for large image collections while being robust to drift due to chaining transformations and minimal overlap between images. We focus on scientific applications where ground truth accuracy is far more important than visual appearance or projection error, which can be misleading. For common largescale image stitching use cases, transformations between images are often restricted to similarity or translation. When homography is used in these cases, the odds of being trapped in a poor local minimum and producing unnatural results increases. Thus, for transformations up to affine, we cast stitching as minimizing reprojection error globally using linear least squares with a few, simple constraints. For homography, we observe that the global affine solution provides better initialization for bundle adjustment compared to an alternative that initializes with a homography-based scaffolding, and at lower computational cost. We evaluate our methods on a very large translation dataset with limited overlap, as well as four drone datasets. We show that our approach is better compared to alternative methods such as MGRAPH in terms of computational cost, scaling to large numbers of images, and robustness to drift. We also contribute ground truth datasets for this endeavor.

PhytoOracle: Scalable, modular phenomics data processing pipelines

“… PhytoOracle two-dimensional (2D) image processing workflow. (A) The 2D pre-processing steps include the conversion of binary (BIN) files (RGB, thermal, PSII chlorophyll fluorescence) to GeoTIFF files, correction of georeferencing information within each GeoTIFF metadata using Megastitch for RGB and thermal data, clipping corrected GeoTIFF images to plots using a GeoJSON file with plot boundary information, and generation of plot level orthomosaics ( Zarei et al., 2022 ). (B) RGB & thermal plot level orthomosaics are run through a Faster R-CNN detection model for plant detection and phenotype extraction; PSII images are run through FLIP for extraction of minimum (F 0 ) and maximum (F M ) fluorescence values, variable fluorescence (F V ), and maximum yield of primary photochemical efficiency (F V /F M ).…”

“…The first container converted BIN files to GeoTIFF images with approximate GPS bounding coordinates calculated from barcode positioning information contained within the JSON metadata file generated by the FS. The second container deployed MegaStitch, which is a software for efficient image stitching of large-scale image datasets ( Zarei et al., 2022 ). Megastitch was run in a non-distributed manner as all images are required for the global optimization stitching method, which generated geometrically corrected GeoTIFFs.…”

“…Thermal calibration measurements were applied to each pixel, converting the DN value to Celsius. The second container deployed MegaStitch ( Zarei et al., 2022 ) in a non-distributed manner, which generated geometrically corrected GeoTIFFs. The third container clipped GeoTIFFs to plot boundaries specified within a GeoJSON file.…”

Geological Borehole Video Image Stitching Method Based on Local Homography Matrix Offset Optimization

“…(2) We can use the homography matrix to calculate the symmetrical transformation error for the rest of the matching point pairs [ 30 , 31 ]. The points, whose values are less than the threshold value, are considered as interior points, as follows: …”