
KS7333
DATA SHEET
10
SYSTEM CONFIGURATION & OPERATION DESCRIPTION
MOTION DETECTION BLOCK
The motion detection block can be divided into the horizontal motion vector detection block and the vertical
motion vector detection block. Its input is the upper 4 bits of the 8-bit luminance signal which is the LPF-handled
part of the 10-bit S1S2 format signal. The block uses the difference between the previous image and the current
image to find the motion vector. To find the motion vector, the current image's luminance value during the input
image's active period must be projected in both horizontal and vertical direction to the current line memory, and
put through correlation matching with the value stored in the previous line memory. In this process, the location
with the smallest correlation error becomes the motion vector. The search for the motion vector is limited to
±
64
in the horizontal direction, and
±
23 in the vertical direction.
To reduce the calculation amount and the time spent in operation, the coarse-to-fine correlation operation is
carried out within the search area. The correlation operation is put into effect within the vertical blank section, and
the motion vector that is finally output has the horizontal value of 7 bits and vertical value of 6 bits.
1-D projection to horizontal/vertical
Coarse-to-fine correlation matching
MSB 4-bit luminance signal input
±
64(H),
±
23(V) search area
Full/Zoom area motion detection according to the zoom ratio
MVX[6:0], MVY[5:0] output
Max, min correlation value output for adaptive image stabilization
ISF BLOCK
The ISF block accumulates the motion vectors (VX, VY) between the image fields to first calculate the integration
value (GX, GY), which is the actual correction value used. If you use the motion vector's integration value, the
motion is corrected flawlessly. However, if the camera user's deliberate movements (panning) are also corrected,
a memory should have compensation limit in image.
To correct such a problem, the accumulated image movement is divided into high frequency and low frequency
components, and only the high frequency components are corrected. To effectively divide these high frequency
components, IIR filtering is independently carried out horizontally and vertically.
At this time, The feedback coefficient of the filter can be selected in MICOM.
10 degree LPF coefficient
Horizontal/Vertical IIR filtering
Temporal filtering output (UX, UY)
Motion vector evaluation (MD_EVAL. V) carried out first
Graphic movement information display (MD_GRAPH. V)