Saliency Detection Using Sparse and Nonlinear Feature Representation

Table 1

Following are the input image resolution used in each algorithm with default parameter settings. Some algorithms (at least*) do internally reduce dimensions for fast computation and optimal performance.

Model: Acronym	Resolution
AIM [16]	1/2(W × H)
*GBVS [37]	W × H
MESR [39]	64 × 48
*ICL [46]	W × H
*Itti [37]	W × H
MPQFT [39]	64 × 48
AWS [23]	W × H
*SDSR [26]	W × H
SUN [47]	1/2(W × H)
HFT [2]	128 × 128
LG [29]	256 × 256
ERDM [17]	512 × 512
ΔQDCT [39]	64 ×48
*Proposed	W × H