Matevž Kovačič and Bojan Kverh and Franc Solina (1997) Optimal models for visual recognition. In: Advances in computer vision. UNSPECIFIED, pp. 51-60.
Minimum description length (MDL) principle can be used as a method for model construction from sensory data. The application of the MDL principle requires the computation of information content of the model. The paper describes the generic encoding for non graph theoretic model. For demonstration, modeling of edges with straight line segments is performed. This example demonstrates that instead of heuristic approaches such low-level image processing tasks can be founded on sound theoretical basis. The issue of time complexity of object reconstruction is briefly addressed. It is proposed that models with various time complexity and accuracy should be used to achieve optimal time complexity along with high reconstruction accuracy.
|Item Type: ||Book Section|
|Keywords: ||computer vision, parametric models, model selection, MDL|
|Language of Content: ||English|
|Institution: ||University of Ljubljana|
|Department: ||Faculty of Computer and Information Science|
|Divisions: ||Faculty of Computer and Information Science > Computer Vision Laboratory|
|Item ID: ||29|
|Date Deposited: ||22 Jan 2003|
|Last Modified: ||13 Aug 2011 00:31|
Actions (login required)