Image Processing Reference
In-Depth Information
Fig. 6.7 Disparity map processing steps (partitioning: 7 regions), “Actors” sequence. (a) YUV
frame. (b) Disparity map. (c) Quantized disparity map. (d) Rescaled disparity map
Fig. 6.8 Disparity map processing steps (partitioning: 6 regions), “Musicians” sequence. (a) YUV
frame. (b) Diparity map. (c) Quantized disparity map. (d) Rescaled disparity map
regions. In many practical cases, the objects that belong to the scene
s foreground
are subject to higher visual attention from the user and are more relevant in terms of
QoE. From this assumption, we can consider that the multi-view disparity map
contains the main information when defining regions of interest. In our case, the
disparity maps are extracted during the post-acquisition step and transmitted with
the corresponding view to the encoder. Then, to improve the compression we will
process this disparity map images in order to obtain a macroblock based
partitioning of each frame as described in Figs.
6.7
and
6.8
.
'
