3DNR Introduction

The three-dimensional noise reduction (3DNR) is used to minimize noise interference in low-signal images. By comparing adjacent frames to automatically filter out non-overlapping information (noise), it produces a cleaner and more refined image. Compared with the noise reduction (NR) module of the ISP, 3DNR performs NR in the YUV domain by combining the temporal and spatial domains, without noise calibration.

3DNR has a hierarchical architecture that provides four levels of filtering. Levels 0 and 3 have spatial filters with temporal assistance, while levels 1 and 2 have both spatial and temporal filters. The chrominance filters are independent of the luminance filters. The overall input of 3DNR includes the current frame and reference frame. The output of each level is the input of the next level, and the reference frame of each level is shared.

Call the hi_mpi_vpss_set_grp_param API to set the 3DNR parameters. For details about the APIs and parameters, see the description of APIs in section " VPSS" .