Geometric registration of images with arbitrarily-shaped local intensity variations from shadows

Abstract

In this paper, we focus on the sub-pixel geometric registration of images with arbitrarily-shaped local intensity variations, particularly due to shadows. Intensity variations tend to degrade the performance of geometric registration, thereby degrading subsequent processing. To handle intensity variations, we propose a model with illumination correction that can handle arbitrarily-shaped regions of local intensity variations. The approach is set in an iterative coarse-to-fine framework with steps to estimate the geometric registration with illumination correction and steps to refine the arbitrarilyshaped local intensity regions. The results show that this model outperforms linear scalar model by a factor of 6.8 in sub-pixel registration accuracy.

Fast computation algorithm of image registration using increment sign correlation and M-estimators

“…K. V. Arya [2] used Huber's estimator and Tukey's bisquare estimator for image registration and showed results for the images having occlusion and shadow. In [3], Fouad focus on the geometric registration of images with disjoint local intensity shifts, especially due to large shadow differences when images captured at different time. Then in [4], Fouad author had extened their work of image registration under illumination variations using region based confidence weighted M -Estimators.…”

High dynamic range imaging of non-static scenes

“…18-20 A recent approach by Fouad et al 21 proposes an affine sub-pixel registration method combined with a segmentation based local illumination model that considers a number of arbitrary linear functions as candidate illumination functions at each pixel. The number of such functions corresponds to the number of possible different illumination levels in two images.…”

Image registration under arbitrarily-shaped local illumination variations

“…First, we propose an intensity-based image registration model that can handle arbitrarily-shaped local illumination variations. We employ the ASLIV model with a LS estimator to accomplish sub-pixel geometric registration jointly with illumination correction as shown in [18]; this approach is 1.3. DISSERTATION OUTLINE 6 referred to LS-ASLIV.…”

“…It is necessary to realize that the LS-ASLIV 6j 2 approach {i.e., "J=l") corresponds to the global illumination approach in [10,11,13]. As well, the LS-ASLiV 6i 4 approach (i.e., "J=2") is a special case discussed in [18] and deals with those image pairs having only two distinct illumination regions.…”

“…2.1(a,b) using the LS-GIM, LS-AIM, LS-ASLIV 6 , 6 and LS-ASLIV 6>8 approaches. These mosaics are simply created using an overlaid average of the two resulting aligned images as presented in [18]. It can be shown that LS-ASLIV 6 , 8 approach yields NCC=0.9384, thereby outperforming both the LS-GIM with NCC=0.8819, LS-AIM with NCC=0.8998 and the LS-ASLIV 6 , 6 with NCC=0.9226.…”