Piecewise Image Registration in the Presence of Multiple Large Motions

Abstract

a) Left image (b) Right image (c) Labeled flow field (d) Reconstructed image Figure 1. Dense pixel correspondence between two images of a scene with moving rigid objects. Images (a) and (b) are the input, and (c) is the flow field from (a) to (b) with color labelings. Red indicates occluded areas, and blue and green indicate two motions modeled by fundamental matrices for the bird and the rest of the background(note that the flow vectors are uniformly scaled to a shorter length for visual clarity). Image (d) is the reconstruction of (a) based on the flow; red areas are occluded. AbstractWe present a technique for computing a dense pixel correspondence between two images of a scene containing multiple large, rigid motions. We model each motion with either a homography (for planar objects) or a fundamental matrix. The various motions in the scene are first extracted by clustering an initial sparse set of correspondences between feature points; we then perform a multi-label graph cut optimization which assigns each pixel to an independent motion and computes its disparity with respect to that motion. We demonstrate our technique on several example scenes and compare our results with previous approaches.

Efficient Image Stitching in the Presence of Dynamic Objects and Structure Misalignment

“…of points to be matched is below some threshold. The above process is similar to the work of [13] and [14]. Different from them, before implementing RANSAC, we remove mismatches by our previous proposed method [15].…”

Simultaneous image de-noising and registration using graph cuts: Application to corrupted medical images

“…Image de-noising was one of the first Graph Cuts applications, and it has recently been shown that Graph Cuts can be used for image registration [15,16,17,18,19,20,21]. We propose a joint approach, where image de-noising helps the registration process, and the good overlap of a clearer image improves de-noising of a corrupted image.…”

Image composition for object pop-out

“…The Photomontage and related work [1,3] are most related to our work, in that they also use a stack of similar images to create a new composite image. However, the main goal of [1] is to create a composition that is visually better than any of the original ones and it has a lot of manual intervention (it is mainly an artist's tool).…”

“…However, the main goal of [1] is to create a composition that is visually better than any of the original ones and it has a lot of manual intervention (it is mainly an artist's tool). [3] uses only one reference image and assumes that the location of input paired images is known. One of our baseline algorithms is an adapted version of [3].…”

“…[3] uses only one reference image and assumes that the location of input paired images is known. One of our baseline algorithms is an adapted version of [3].…”