A novel high performance discrete cosine transform algorithm based on the stochastic computation
Qian Wang, Xiao Guoqiang, Lin Xiao · 2013
This paper proposed a novel high performance Discrete Cosine Transform (DCT) algorithm which is implemented by stochastic computation. Firstly, the image information represented in the format of binary radix is mapped to the probability field in which the numerical value is represented by a stochastic bits stream. Then, a series of calculation rules based on stochastic computation theory are established for the implementation of DCT and IDCT. For example, one AND gate multiplies two values in probability field. Finally, the performance is tested by PSNR (Peak Signal to Noise Ratio) between the images reconstructed by DCT and IDCT algorithm with the stochastic calculation and floating-point calculation. Experimental result illustrates that the introduced approach is simply-constructed which possesses a low logic cost and high performance .