Sub-quantization/orthogonalization and optimization of algorithm-architecture adequacy for optimal polynomial filtering

This paper deals with the benefits of mixing an orthogonalization procedure to the sub-quantization of the input operation, in order to reduce the complexity and to enhance the robustness implementation of optimal polynomial filters. These non linear filters are now generic devices for real-time high speed multimedia applications. We propose an optimal polynomial filtering scheme based on Sub-quantization/ Orthogonalization operations, related to the current interest to complexity, power consumption and areas reduction. The orthogonal polynomial basis is chosen to be in a simple closed-form expression and the input is sub-quantized with a desired sub-quantization degree. We explore, from stability and accuracy tests, the suitable choice of sub-quantization in order to achieve performances of the proposed optimal polynomial filtering scheme with robustness performances almost similar to the corresponding continuous input case.