RAVA: Region-Based Average Video Quality Assessment

. 2021 Aug 15;21(16):5489. doi: 10.3390/s21165489

Algorithm 1: Proposed RAVA algorithm.

RAVA( $I, M$ ): I represents list of all the frames of a video, and M is the number of frames.

1:
2:
$i \leftarrow 1$
3:
for all $i \leftarrow 1 t o M - 1$ do
4:
$B b o x e s_{i} \leftarrow l o c a t e o b j e c t s i n f r a m e I_{i}$
5:
$B G_{i} \leftarrow I_{i}$
6:
for all $k \leftarrow 1 t o l e n (B b o x e s_{i})$ do
7:
$R_{i, k} \leftarrow B b o x e s_{i} [k]$
8:
$B G_{i} \leftarrow I_{i} ⊖ R_{i, k}$ {Mask out foreground regions}
9:
$m a g_{i, k}, α_{i, k} \leftarrow O p t i c a l f l o w (R_{i, k} \to R_{i + 1, k})$
10:
$a v g_{-} o p t_{i, k} \leftarrow \frac{\sum_{j = 1}^{K} m a g_{i, k, j}}{K}$
11:
{K is the number of pixels in this region $R_{i, k}$ }
12:
end for
13:
end for
14:
for all $i \leftarrow 1 t o M - 1$ do
15:
for all $k \leftarrow 1 t o N$ do
16:
{N is the number of regions in the frame}
17:
$w_{i, k} = \frac{a v g_{-} o p t_{i, k}}{\sum_{m = 1}^{M - 1} \sum_{n = 1}^{N} a v s . g_{-} o p t_{m, n}}$
18:
19:
end for
20:
end for
21:
$F G f e a t u r e = \frac{\sum_{m = 1}^{M - 1} \sum_{n = 1}^{N} w_{m, n} \cdot I Q A (R_{m, n})}{\sum_{m = 1}^{M - 1} \sum_{n = 1}^{N} 1}$
22:
23:
$B G f e a t u r e = \frac{\sum_{m = 1}^{M} I Q A (B G_{i})}{M}$
24:
{IQA can be PSNR or SSIM}