Application Programs and Systems - Heterogeneous Multicore Processor Technologies for Embedded Systems - page 210

First, the MPEG-2 decoding is processed on the VPU using a frame buffer of
decoding data, whose size corresponds to four frames of the video image. The VPU
starts the decoding frame-by-frame when one frame of an input data stream is
obtained from the memory, and it stores the decoded image to one of the four frames
in the frame buffer.
The still-image capturing duplicates the decoded image to the frame buffer of the
captured image at every decoding frame. The buffer of the captured image is shared
with uITRON and Linux in the memory; therefore, a program on Linux can obtain
a decoded image any time.
The image scaling also duplicates the decoded image to the frame buffer of the
decoded data at every decoding frame. Since the adjusted size of images at the scaling is
set to 720 × 480, scaling factors in both horizontal and vertical directions to the decoded
image are calculated and set to the VEU. For example, when the size of the image is
720 × 480, the scaling factors are set to 1.00 and 1.00 in the horizontal and vertical direc-
tions, respectively. In the same manner, when the size is 960 × 540, the scaling factors are
set to 0.75 and 0.89. When the size is 320 × 240, the factors are 2.25 and 2.00. After the
start-up of the VEU, it reads an image from the frame buffer of the decoded data, adjusts
the size of the image according to specified scaling factors, and writes the scaled image
whose size is 720 × 480 to the frame buffer of the scaled data.
The video image and graphics synthesizing process uses image data in the frame
buffer of the scaled data, as well as graphics data in the frame buffer of graphics and
blends them in the BEU. The size of the frame buffers is 1,024 × 768. When a scaled
image is stored in the frame buffer of scaled data, the BEU starts the blending and
writes the synthesized image to the frame buffer of the synthesized data. The graphics
frame buffer is placed in the memory area shared by both uITRON and Linux and
can therefore be updated on Linux at any time.
Finally, the output image control sets up the LCDC and a DVI transmitter to con-
vert the synthesized image stored in the frame buffer into video signals that are trans-
mitted to the monitor via the DVI interface. Figure 6.38 illustrates the processing
flow of the uITRON system. The process is repeated from supplying the video stream
to copying the frame buffer of decoding data to that of a still-captured image.

6.4.2 Implementation of Face Recognition and GUI Controls

The system on Linux performs face recognition by utilizing the similar-image
search, pointing device detection, and GUI controls to create a graphics plane gen-
erated by the face recognition. Figure 6.39 depicts a block diagram of the Linux
system that comprises the following five functions:

1. Similar-image search
2. Face detection
3. Event processing
4. Image object management
5. Image processing