Image Pulse Coding Scheme applied to Feature Extraction

The resurgence in interest concerning pulse-coded neural nets has led to various attempts to integrate pulse computation schemes in conventional mixed signal VLSI. Of interest is especially the robustness of representing an analogue signal in the phase or duration of a pulsed, quasi-digital signal, as well as the possibility of direct digital manipulation of such an analogue signal. Starting from the biological model, various pulse coding and computation schemes have been postulated, such as rate codes with different rate definitions, absolute or relative phase codes, correlation codes, various single or multi-neuron statistical codes, rank order coding, or neural noise shaping. Simply using pulse order, we will show that a powerful low-level feature and texture extraction operator, called Local Orientation Coding, can be implemented efficiently as a pulsed computation scheme. Feature extraction results are being presented, and possible VLSI implementations are detailed.