In some embodiments, instead of using a high pass filter (e.g., 214-2 of FIG. 3A or 210-2 of FIG. 3B) in a demultiplexer (160-2 or 160-3) as illustrated in FIG. 3A and FIG. 3B to derive full resolution LE image frame (e.g., 162-L), a subtraction operation as illustrated in FIG. 4A may be used. In contrast to high pass filtering operations, a subtraction operation such as 402 may require less computational complexity.
For the purpose of illustration only, the first LE portion (202-L) may be decoded from a BL video signal, while the second LE portion (204-L) may be decoded from an EL video signal. As illustrated in FIG. 4A, a reference LE portion 404, which may be of the same spatial dimensions of the second LE portion (204-L), may be generated based on the first LE portion (202-L) by RPU 406 (which may be, for example, 154 of FIG. 1B) during an inter-layer prediction process. In some embodiments, a processing path or sub-path comprising a low pass filter (408) in the vertical direction (which removes high spatial frequency content in the vertical direction), a vertical subsampler (410; which may keep, for example, every other row), and a horizontal up-sampler (412) may be used to generate the reference LE portion 404.
