This paper formulates the visual object recognition problem in the discriminant analysis framework and presents a kernelized version of the transformational approach of distance-based discriminant analysis. The sought transformation is found as a solution to an optimization problem formulated in terms of inter-observation distances only, using the technique of iterative majorization. The proposed approach is non-parametric, and can determine the dimensionality of the target space automatically since the process of feature extraction is fully embedded in the optimization procedure. Performance tests and experiments in the application of visual object and content-based image categorization demonstrate very competitive results in comparison to several popular existing techniques.