Color Image Processing
Color based object Recognition
The research topic is color based object recognition by using specific object information with methods for localizing and identification. For this purpose, an image database contains several objects taken under different illuminations and orientations. An ideal recognition method is determined regarding the selected object during the training. In this phase the position of the objects within the images of the training set is known. This leads to a processing chain adapted to the specific object which gives the best recognition results. The chain consists of noise filters, algorithms of color constancy, color space conversions, histogram methods, etc. The final process should work for all objects in all scenes with unknown illumination and orientation using the determined chain from the training phase and a known query object.
The results obtained by the object recognition task should be examined additionally in terms of knowing the image formation process. There it needs to be resolved in how far the (spectral) camera response, illumination, and object reflectances, influence the recognition rates?