Geometric invariance in computer vision
Geometric invariance in computer vision
Fundamentals of neural networks: architectures, algorithms, and applications
Fundamentals of neural networks: architectures, algorithms, and applications
Applied image processing
Pattern recognition and image analysis
Pattern recognition and image analysis
| Hi-index | 0.02 |
In this paper we evaluate the performance of self-organising maps (SOM) for image classification using invariant features based on run length alone and also on run length plus run length totals, for horizontal runs. Objects were manually separated from an experimental set of natural images. Object classification performance was evaluated by comparing the SOM classifications independently with a manual classification for both of the feature extraction methods. The experimental results showed that image classification using the run length method that included run length totals achieved a recognition rate that was, on average, 4.65 percentage points higher that the recognition rate achieved with the normal run length method. Thus the extended method is promising for practical applications.