Attribute Profiles Without Thresholds Through Histogram Based Tree Path Description

Classification of Remote Sensing Data With Morphological Attribute Profiles: A decade of advances

“…More recently, unsupervised threshold selection methods [64], [73] and threshold-free APs [66], [74], [75] have been proposed. In [73], from the AP computed from all possible threshold values of an attribute, a genetic algorithm selects the subset of reconstructed images which convey the highest amount of information.…”

“…Then, a more efficient threshold selection method is proposed in [64]: first, the set of attribute values of all nodes of a component tree are sorted in non-decreasing order; then, an increasing curve is built from the sorted attribute values; finally, the points in the curve with the largest gradient values are the selected threshold values. In [74], [75] a threshold-free approach was introduced, where every tree node is described based on simple statistical properties of the sequence of attribute values belonging to the nodes in the path connecting the node under study to the root of the tree. Properties such as the highest change of attribute value have shown promise.…”

“…First, apart from threshold-free APs [66], [74], [75], the quality of an AP depends on the selected set of thresholds. A bad selection of thresholds can lead to redundant information in the AP [61], [62].…”