3DNR Technology (3D Noise Reduction) is a method of suppressing noise in an image, appearing in low light.
In video transmission systems which also include the video surveillance systems, a special place is given to noise filtering algorithms. Noise reduction is crucial for the overall functioning of the system, since the presence of noise in the video signal not only degrades the image quality, but also affects the subsequent processes of signal processing. Noise is especially harmful to the digital video that is compressed and then decompressed.
Currently, methods of noise reduction can be divided into two types: 2-dimensional 2DNR, (divided then into spatial and temporal) and 3-dimensional 3DNR noise reduction.
The spatial noise reduction filter analyzes the picture only in the spatial domain, ignoring the information in the temporal direction. Temporal noise reduction filters analyze only the pixels in the temporal direction. For the temporal noise reduction, adaptive or compensatory techniques can be used. With the usage of the adaptive method the pixels that are in the same position in different frames are analyzed. Compensatory technique is based on the analysis of trajectories, based on the factual evidence received from the evaluation of movement. But 2DNR method has the disadvantage – when processing the signal, the image details become blurred. 3DNR noise reduction filter combines the advantages of temporal filters with spatial filters, but at the same time has no their disadvantages.
For 3DNR noise reduction, the method of the reduction of the additive influence of the Gaussian noise analyzing the set of consecutive video frames with the temporal filtering is used. The method determines the degree of difference between the pixels in the current frame and pixels in the previous frame. It also defines the motion vector indicative for the movement of pixels in the current frame, and a similar motion of the compensated pixel in the filtered frame. The method then estimates the distortion that affects the pixel in the current frame. As a result, the filter calculates the result at the average “weight” of pixels in the current frame in view of pixels of the second frame, taking into account the results of detection and motion estimation, motion compensation and estimation of noise.
With this method, you can get a quality video signal image under adverse lighting conditions.