- Definitions of global and local visual features
- Visual classification
- Early fusion of visual features: reduction of time computation
- Results of early fusion of visual features

Classification using visual features

To calculate the visual distance between an image of and an image of , we calculate the possible distances and we calculate the average of the smallest values ( ). We obtain for each image the distance:

Now, if one considers the distances between an image of , and all images contained in a class of , one can calculate the final distance between and averaging only the first minimal distances. Then we have:

where is an element of the class of the base of examples and is the number of minimal values taken among the distances. Again the class of considering feature is given by:

This method allows to reject the too large distances which would penalize the system, and to keep the best distances which increases the probability of being in the good class.

It is also noticed that, for (4+g), the global indices make a clear improvement of the ER, except in the case of the direction feature, which was foreseeable. If one compares these results with those of table 3, one notices a fall of about 5% to 10% of ER using the local indices, and an improvement of 2% on the global ones. Moreover, the early fusion reduces the time computation.