Non-rigid consistent registration of 2D image sequences.

We present a novel algorithm for the registration of 2D image sequences that combines the principles of multiresolution B-spline-based elastic registration and those of bidirectional consistent registration. In our method, consecutive triples of images are iteratively registered to gradually extend the information through the set of images of the entire sequence. The intermediate results are reused for the registration of the following triple. We choose to interpolate the images and model the deformation fields using B-spline multiresolution pyramids. Novel boundary conditions are introduced to better characterize the deformations at the boundaries. In the experimental section, we quantitatively show that our method recovers from barrel/pincushion and fish-eye deformations with subpixel error. Moreover, it is more robust against outliers--occasional strong noise and large rotations--than the state-of-the-art methods. Finally, we show that our method can be used to realign series of histological serial sections, which are often heavily distorted due to folding and tearing of the tissues.

Automatic registration of serial mammary gland sections.

We present two methods for automatic registration of microscope images of consecutive tissue sections. They represent two possibilities for the first step in the 3-D reconstruction of histological structures from serially sectioned tissue blocks. The goal is to accurately align the sections in order to place every relevant shape contained in each image in front of its corresponding shape in the following section before detecting the structures of interest and rendering them in 3D. This is accomplished by finding the best rigid body transformation (translation and rotation) of the image being registered by maximizing a matching function based on the image content correlation. The first method makes use of the entire image information, whereas the second one uses only the information located at specific sites, as determined by the segmentation of the most relevant tissue structures. To reduce computing time, we use a multiresolution pyramidal approach that reaches the best registration transformation in increasing resolution steps. In each step, a subsampled version of the images is used. Both methods rely on a binary image which is a thresholded version of the Sobel gradients of the image (first method) or a set of boundaries manually or automatically obtained that define important histological structures of the sections. Then distance-transform of the binary image is computed. A proximity function is then calculated between the distance image of the image being registered and that of the reference image. The transformation providing a maximum of the proximity function is then used as the starting point of the following step. This is iterated until the registration error lies below a minimum value.