Image registration: Features and applications

Abstract

Nowadays images provide more and more information about this world. Often multiple images share the same scene observed from different angles, at different times or with different devices. Image registration is a method of aligning two or more images of the same scene into the same coordinate system so that the aligned images can be directly compared and combined. It is a fundamental step in many image analysis tasks in which the final knowledge has to be gained from the combination of multiple data sources. Identifying the correspondence between two images is simple for human visual system but challenging for computer algorithms. In general, four components are important for a typical image registration framework: image feature extraction, similarity metric, transformation model and optimization strategy. Due to the variety of image types and application domains, it is impossible to design a universal method for all image registration tasks.

In this thesis, we have developed several contributions to the field of image registration. These contributions stand on their own as valuable components within their particular application domains, but are linked under the common theme of image registration. First, we have developed a method which is capable of estimating the skew distortion and orientation of printed document images. It registers a skewed document image with an imaginary image that would be captured if the document was posed in exactly upright position during the scanning procedure. Within this method, we have presented a novel image feature called interline white run to perform this registration task. Interline white run can be accurately derived from white run histograms which are obtained through one-time fast scanning of the document. Although the new feature seems simple, our experiments on real-world documents have demonstrated its efficiency in estimating the skew angle of printed document images.

We have also developed a framework to register the two sides of a double-sided historical document. As historical document images are usually degraded by various noises and distortions, we have designed an algorithm to extract salient control points from historical images for the purpose of registration. For documents with slight geometric distortions, a representative block is selected and used to estimate a rigid transformation model. When severe local deformation is present, mainly warping effects and local uneven surfaces, a fine registration procedure which combines salient points extraction, free-form transformation model and residual complexity similarity measure is additionally applied. Our experiments have shown that this registration framework significantly improves the performances of subsequent bleed-through correction methods.

Finally, we have proposed a groupwise image registration framework to build a brain CT atlas with the CT scans of multiple patients. The groupwise registration method is built upon a non-rigid pairwise image registration method which shares the same transformation model with the method we have proposed for historical document images. CT slices which are from normal study cases and labeled with the same level number are first clustered into different groups. Among each group, all slices are registered to the center of the group and an intermediate average slice is computed for the group. The final average slice for a particular level is the combination of the average slices of all groups on this level. With the built atlas, we can efficiently estimate the level of an input CT slice in the axial direction of brain, which will significantly speed up subsequent content based retrieval systems. In addition, by comparing the input slice which are affected by traumatic brain injury against the atlas, we can identify the abnormal regions on the input slice.