Real-time data-driven systems often utilize discrete valued time series data and their functionality is highly dependent on the accuracy of such data. In order to improve the performance of these systems, an important pre-processing step is the denoising of data before performing any action (e.g. forecasting or control activities). Existing algorithms have primarily focused on the offline denoising problem, which requires the entire data to be collected before the denoising process. In this paper, the problem of online discrete denoising is considered. The online denoising problem is motivated by real-time applications, where the data must be utilizable soon after it is collected. Three online denoising algorithms are proposed which can strike a tradeoff between delay and accuracy of denoising. It is also shown that the proposed online algorithms asymptotically converge to a class of optimal offline block denoisers.