Visual object recognition is performed without effort by humans even though it requires a series of complex computations which are, for now, not well understood. This study relies on the concept of minimal images, smallest configurations where an image is recognizable to the human vision, to study the processes by which the brain uses visual features to carry out computations underlying visual recognition. The role of these visual features is revealed at the minimal level and a tiny change in the image configuration is enough to completely lose recognition. A neurophysiological experiment was conducted with twelve subjects implanted with intracranial electrodes.