الفهرس 
Title : Learning-based feature super-resolution for low-resolution image classification
SummaryOnly 14 pages are availabe for public view
 |  | from | 113 |  |  |
| | | from | 113 | | |
Abstract
The classification of images from their visual texture has many applications ranging from medical diagnosis applications to image retrieval and object recognition. As image resolution determines the amount of details an image holds, it plays an important role when using digital images for classification tasks. The problem we address in this thesis is one of automatically classifying textural images with low resolution conditions since high resolution images are not always available. In this work, we propose learning-based approaches to infer high-resolution features from low-resolution features extracted from low-resolution images. Applying these learned maps is equivalent to super-resolution (SR) in the feature domain. Two different applications are studied in this work. Experimental and statistical evaluations show significant improvement in classification performance due to applying the proposed techniques in comparison with direct classification in the low-resolution space