Optimising the Complete Image Feature Extraction Chain

The hypothesis veri cation stage of the traditional image processing approach, consisting of low, medium, and high level processing, will su er if the set of low level features extracted are of poor quality. We investigate the optimisation of the feature extraction chain by using Genetic Algorithms. The tness function is a performance measure which re ects the quality of an extracted set of features. We will present some results and compare them with a Hill-Climbing approach.

• Type: Conference
• Year: 1998
• Where: ACCV
• Authors: Majid Mirmehdi, Phil L. Palmer, Josef Kittler
• Keywords: ACCV 1998, Computer Vision, Hypothesis Veri Cation, Image Processing Approach, Low Level Features