Scene analysis is an important area of research with the aim of identifying objects and their relationships in natural scenes. MINERVA benchmark has been recently introduced in this area for testing different image processing and classification schemes. In this paper we present results on the classification of eight natural objects in the complete set of 448 natural images using neural networks. An exhaustive set of experiments with this benchmark has been conducted using four different segmentation methods and five texture-based feature extraction methods. The results in this paper show the performance of a neural network classifier on a ten fold cross-validation task. On the basis of the results produced, we are able to rank how well different image segmentation algorithms are suited to the task of region of interest identification in these images, and we also see how well texture extraction algorithms rank on the basis of classification results.

Neural networks have been widely regarded as powerful classifiers. They are particularly suited in scene analysis applicatio n areas where the task is to correctly label different image objects in a scene. Only after this stage is successful, one can build upon this information to generate a semantic understanding of the complete scene. In this paper we investigate the application of neural networks to the classification of natural objects in natural scenes available as a part of the publicly available MINERVA benchmark. The image processing component consists of image segmentation procedure that generates region definitions in images and texture algorithms that compute characteristic features from these. Our results are based on a total of 40 data sets generated using a combination of four segmentation algorithms and five grey scale texture extraction algorithms for classifying vegetation classes and natural object classes. In addition, on the best segmentation method, colour texture features based on correlograms and colour moments are investigated. The paper presents exhaustive results on these data sets and compares the utility of neural networks on a ten-fold cross-validation task.