Citation Context

.... These methods rely on accurate registration of the candidate characters i.e. framing and normalisation as seen in the previous section. This set of experiments uses the PCA classification algorithm =-=[8]-=-. Generally OCR is applied to a single image only, but in this work multiple instances of the same character exist across a number of frames. This fact is exploited by our algorithms. The differing vi...

Citation Context

...on of arbitrary regions based on size, colour, intensity, texture and neighbourhood is more tractable making it suitable for tracking [1]. Pre-segmentation methods [2] such as the watershed transform =-=[3]-=-[4], and clustering [5], can result in over segmented graphs, and efficient merging algorithms are required to join adjacent and similar regions, which can be time consuming. A region adjacency graph ...

Citation Context

...l scenes using knowledge of the clothing colour. The isolated player is then more closely scrutinised and the region containing the number is extracted. Optical Character Recognition (OCR) techniques =-=[7]-=- are applied to the candidate number regions, and from it the player identity derived. The recognised number can be indexed in a look-up table and used for the provision of augmented information. Repo...

Citation Context

...scene is difficult, the identification of arbitrary regions based on size, colour, intensity, texture and neighbourhood is more tractable making it suitable for tracking [1]. Pre-segmentation methods =-=[2]-=- such as the watershed transform [3][4], and clustering [5], can result in over segmented graphs, and efficient merging algorithms are required to join adjacent and similar regions, which can be time ...

Citation Context

...[4], and clustering [5], can result in over segmented graphs, and efficient merging algorithms are required to join adjacent and similar regions, which can be time consuming. A region adjacency graph =-=[6]-=- (RAG), provides a simple-connectivity view of the image and facilitates merging. The adjacent regions can be merged provided they have sufficiently close colour distribution. Picture trees can be for...

Citation Context

...of arbitrary regions based on size, colour, intensity, texture and neighbourhood is more tractable making it suitable for tracking [1]. Pre-segmentation methods [2] such as the watershed transform [3]=-=[4]-=-, and clustering [5], can result in over segmented graphs, and efficient merging algorithms are required to join adjacent and similar regions, which can be time consuming. A region adjacency graph [6]...

Citation Context

... based on size, colour, intensity, texture and neighbourhood is more tractable making it suitable for tracking [1]. Pre-segmentation methods [2] such as the watershed transform [3][4], and clustering =-=[5]-=-, can result in over segmented graphs, and efficient merging algorithms are required to join adjacent and similar regions, which can be time consuming. A region adjacency graph [6] (RAG), provides a s...

Citation Context

...ion of generic objects from a scene is difficult, the identification of arbitrary regions based on size, colour, intensity, texture and neighbourhood is more tractable making it suitable for tracking =-=[1]-=-. Pre-segmentation methods [2] such as the watershed transform [3][4], and clustering [5], can result in over segmented graphs, and efficient merging algorithms are required to join adjacent and simil...