Fuzzy motion adaptive algorithm and its hardware implementation for video de-interlacing

Interlacing techniques were introduced in the early analog TV transmission systems as an efficient mechanism capable of halving the video bandwidth. Currently, interlacing is also used by some modern digital TV transmission systems, however, there is a problem at the receiver side since the majority of modern display devices require a progressive scanning. De-interlacing algorithms convert an interlaced video signal into a progressive one by performing interpolation. To achieve good de-interlacing results, dynamical and local image features should be considered. The gradual adaptation of the de-interlacing technique as a function of the level of motion detected in each pixel is a powerful method that can be carried out by means of fuzzy inference. The starting point of our study is an algorithm that uses a fuzzy inference system to evaluate motion locally (FMA algorithm). Our approach is based on convolution techniques to process a fuzzy rulebase for motion-adaptive de-interlacing. Different strategies based on bi-dimensional convolution techniques are proposed. In particular, the algorithm called ‘single convolution algorithm’ introduces significant advantages: a more accurate measurement of the level of motion by using a matrix of weights, and a unique fuzzification process after the global estimation, which reduces the computational cost. Different architectures for the hardware implementation of this algorithm are described in VHDL language. The physical realization is carried out on a RC100 Celoxica FPGA development board.