Image processing filter selection by block design and its simple case study

The performance of feature extraction strongly depends on the image filters employed. However, it is difficult to select an optimal set of filters for a given series of images. The present paper proposes an approach for choosing an optimal set of filters, given a set of images. In the proposed method, the optimal filter set is refined by comparing the feature extraction performances for a sample image from the image set. The proposed method uses a block design to reduce the number of comparisons. We also discuss the application for surveillance camera as a simple case study.