Figure 1.
Saliency maps vs. ground truth.
Given several original images [20] (top), our saliency detection method is used to generate saliency maps by measuring regional principal color contrasts (middle), which are comparable to manually labeled ground truth [11] (bottom).
Figure 2.
Color space distribution and quantization.
(A) Input image [20]. (B) Original color distribution of A in the RGB color space. (C) Color distribution of uniform quantization. (D) Color distribution of minimum variance quantization.
Figure 3.
Replacement for low frequent colors with minimum variance quantization.
(A) Minimum variance quantized Fig. 2A. (B) Color histogram of the image in A. (C) Full resolution output image resulting from the retained high frequent colors. (D) Color histogram of C.
Figure 4.
Saliency map generated by global color contrast.
(A) Global color saliency. (B) Color space smoothing.
Figure 5.
Regional principal color contrast.
(A) Regional boundaries of using the graph-based segmentation method. (B) Each region represented by its principal color. (C) Saliency map obtained with the saliency values of regional principal colors.
Figure 6.
Saliency map with measuring two categories of spatial relationships.
(A) Between two regions. (B) Between regional center and image center. (C) Binary segmented result simply obtained by thresholding B with an adaptive threshold.
Figure 7.
Visual results of our method compared with ground truth and other methods on dataset MSRA-1000.
(A) Original images [20]. (B) Ground truth [11]. (C) IT [1]. (D) SR [14]. (E) FT [11]. (F) CA [19]. (G) RC [10]. (H) Ours.
Figure 8.
Quantitative comparison on dataset MSRA-1000 (N/A represents no center-bias).
(A) Precision-Recall curves. (B) F-measure curves. (C) Precision-Recall bars.
Table 1.
Numeric comparison on data set MSRA-1000 (%, N/A represents without center-bias).
Figure 9.
Hard image cases of our method in detecting salient regions.
(Top to bottom) Original images [20], ground truth (GT) [11], color histogram similar to Fig. 3D, global color contrast, and regional principal color based saliency detection.
Figure 10.
Uniform quantization vs. minimum variance quantization.
(A) Precision-Recall curves. (B) Precision-Recall bars. (C) F-measure curves.
Table 2.
Numeric comparison for various colors used in uniform quantization (%).
Table 3.
Numeric comparison for various colors in minimum variance quantization (%).
Figure 11.
Quantitative comparison for various combinations of parameters.
(A)–(C) Varying σ from 0.01 to 1 with α = 0.95 and δ = 1/4: (A) Precision-Recall curves. (B) F-measure curves. (C) Ps Bars. (D)–(F) Plots of precision, recall, and F-measure for various values of σ: (D) α = 0.9, δ = 1/16 vs. α = 0.95, δ = 1/16. (E) α = 0.9, δ = 1/4 vs. α = 0.95, δ = 1/4. (F) α = 0.95, δ = 1/16 vs. α = 0.95, δ = 1/4.
Table 4.
Numeric comparison for various combinations of parameters (%).