VS6724 Datasheet, PDF (16/118 Pages) STMicroelectronics – 2 Megapixel single-chip camera module

Functional description
VS6724
2.4.9
2.4.10
2.4.11
Defect correction 1 & 2
The input to the defect correction filters is a 5x5 pixel matrix from which a centre pixel and a
neighborhood of 8 pixels is extracted. The 8 pixel neighborhood is used to determine the
validity of the corresponding centre pixel. This is achieved using the ranked output of a
Batcher-Banyan sort architecture. The defect correction filters perform both detection and
weighted correction on all data. Defect correction 1 is weighted to work horizontally and
vertically. Defect correction 2 is weighted to operate on the diagonal elements.
Spatial noise reduction
The noise reduction module implements an algorithm based on the human-visual system
and adaptive pixel filtering that reduces perceived noise in an image whilst maintaining
areas of high definition.
Interpolation
The interpolation module converts Bayer pixel data to RGB and applies an anti-alias filter to
the data. It also generates the sharp data used in the aperture correction block.
2.5
2.5.1
Note:
Color engine (CE block)
The role of the CE is to process the RGB data from the recovery engine into a specific
format requested by the host, e.g. RGB565 for LCD display, or YCbCr for the JPEG encoder.
Color matrix
A matrix color correction transform is performed on the outputs of the scaler, the matrix is
detailed in Figure 6.
Figure 6. Color matrix
rcof00
rcof10
rcof20
rcof01
rcof11
rcof21
rcof02
rcof12
rcof22
The matrix equations are:
Ro = Rin.rcof00 + Gin.rcof01 + Bin.rcof02
Go = Rin.rcof10 + Gin.rcof11 + Bin.rcof12
Bo = Rin.rcof20 + Gin.rcof21 + Bin.rcof22
The neutral-preserving (“eigen-grey”) property of the matrix can be achieved by ensuring
the matrix rows sum to unity.
16/118