Video displays have started with a grey-scale CRT based systems and have evolved ever since. The displays of today provide true to life and large display of images and video, giving the viewer a gripping experience. Displays like Samsung’s 102" plasma display or the Phillips 3D TVs are some of the examples for the advancement in the video displays. Such large displays need very clean video content to maintain the value of the product.

Noise is inherent in all the data and communication channels. Even in videos, the noise can be added to the videos at different places from the time it is captured to the time it is displayed. Since the video dimensions fixed to a particular standard, the displays use scalars to fit them to size of the screen. This makes the fine noise present in the videos more visible and degrades the visual quality of the displayed video. Hence denoising of videos becomes a critical task.

The video noise can be broadly classified into as Gaussian noise and compression noise. Gaussian noise is due to the combination of various noise types added to the input video from the time of capture to the time it is displayed. Compression noise, as the name suggests is introduced in videos due to compression. Many a times, the videos are compressed to meet the bandwidth limitations for applications like video on demand or streaming video. This leads to artifacts like the ringing and blocking in the videos that attribute to compression noise.

The objective of this project is to develop algorithms to clean the noisy videos while preserving the details to a large extent. These algorithms must be for real time processing and hardware friendly.

This report discusses algorithms developed to reduce noise in the videos. The noise is classified into Gaussian and compression noises. The compression noise is further divided into ringing and blocking artifacts. The characteristics of the noise and algorithms to reduce it are discussed in the report.