In this work, we demonstrate a systematic way to analyze a binary partition tree representation of natural images for the purposes of archiving and segmentation. Within the tree structure, these problems are transformed into locating prevalent tree branches. With a user interface these points can be found manually by browsing branches. However, tree visualization is difficult due to the high node density. A simpler version of the tree is desired which facilitates subsequent retrieval whilst maintaining as much semantic detail as possible. By studying the evolution of region statistics, our method highlights nodes which represent the boundary between salient detail and provide a set of tree levels from which simplifications and segmentations can be derived. A series of subjective tests are performed to demonstrate the effectiveness of using the simplified trees for object extraction. Segmentation results are compared to ground truths showing semantic content is maintained

Pyramid image representations via tree structures are recognized methods for region-based image analysis. Binary partition trees can be applied which document the merging process with small details found at the bottom levels and larger ones close to the root. Hindsight of the merging process is stored within the tree structure and provides the change histories of an image property from the leaf to the root node. In this work, the change histories are modelled by evolvement functions and their second order statistics are analyzed by using a knee function. Knee values show the reluctancy of each merge. We have systematically formulated these findings to provide a novel framework for binary partition tree analysis, where tree simplification is demonstrated. Based on an evolvement function, for each upward path in a tree, the tree node associated with the first reluctant merge is considered as a pruning candidate. The result is a simplified version providing a reduced solution space and still complying with the definition of a binary tree. The experiments show that image details are preserved whilst the number of nodes is dramatically reduced. An image filtering tool also results which preserves object boundaries and has applications for segmentation.

Tracking objects in motion is often done by imposing the constraints of kinematics and local image properties onto the objects. In this work, we propose a novel tracking algorithm which uses the surrounding information of the object to construct the feature profiles. The object feature profiles are then compared across consecutive frames to locate the targets. The feature profiles possess two important properties, distinctive-ness and coherence, which make them robust to measurement noises, short occlusions and false targets. The matching cost function is formulated under a Bayesian framework that enables the algorithm to capture the properties in the form of probabilities. The algorithm is also self-initializing. The computation of the feature profiles is fast due to their simple definition; and the comparison between two profiles can also be done efficiently.