2nd psychophysical experiment - ours against other methods

• This is the comparison of the proposed method against the original, Matlab's histogram equalization, and Photoshop's auto contrast, respectively.

• Each image has been tested for two different keywords.

• Approval rates from 40 observers are indicated in brackets.

• Due to the 10MB limitation, images are downsampled and jpg-compressed.

Some examples we want to point out

• Our semantic enhancement is preferred for both semantic concepts -- dark and snow.

• Our semantic enhancement is NOT preferred for light, but for silhouette.

• Our semantic enhancement is preferred for sand, but NOT for silhouette.

• Our semantic enhancement is preferred for both semantic concepts -- sand and silhouette.

keyword = dark
original (26%)	histogram equalization (0%)	Photoshop auto contrast (8%)	proposed (67%)
keyword = snow
original (5%)	histogram equalization (65%)	Photoshop auto contrast (11%)	proposed (19%)

keyword = dark
original (16%)	histogram equalization (0%)	Photoshop auto contrast (3%)	proposed (81%)
keyword = snow
original (0%)	histogram equalization (44%)	Photoshop auto contrast (3%)	proposed (53%)

keyword = dark
original (14%)	histogram equalization (0%)	Photoshop auto contrast (8%)	proposed (78%)
keyword = snow
original (3%)	histogram equalization (59%)	Photoshop auto contrast (3%)	proposed (35%)

keyword = dark
original (8%)	histogram equalization (0%)	Photoshop auto contrast (19%)	proposed (72%)
keyword = sand
original (3%)	histogram equalization (38%)	Photoshop auto contrast (0%)	proposed (59%)

keyword = dark
original (11%)	histogram equalization (0%)	Photoshop auto contrast (5%)	proposed (84%)
keyword = sand
original (6%)	histogram equalization (25%)	Photoshop auto contrast (14%)	proposed (56%)

keyword = dark
original (0%)	histogram equalization (22%)	Photoshop auto contrast (3%)	proposed (75%)
keyword = sand
original (13%)	histogram equalization (11%)	Photoshop auto contrast (8%)	proposed (68%)

keyword = dark
original (8%)	histogram equalization (0%)	Photoshop auto contrast (5%)	proposed (86%)
keyword = snow
original (11%)	histogram equalization (28%)	Photoshop auto contrast (6%)	proposed (56%)

keyword = dark
original (3%)	histogram equalization (0%)	Photoshop auto contrast (15%)	proposed (82%)
keyword = snow
original (3%)	histogram equalization (50%)	Photoshop auto contrast (11%)	proposed (37%)

keyword = dark
original (16%)	histogram equalization (5%)	Photoshop auto contrast (11%)	proposed (68%)
keyword = snow
original (3%)	histogram equalization (38%)	Photoshop auto contrast (0%)	proposed (59%)

keyword = light
original (5%)	histogram equalization (53%)	Photoshop auto contrast (29%)	proposed (13%)
keyword = silhouette
original (13%)	histogram equalization (0%)	Photoshop auto contrast (37%)	proposed (50%)

keyword = sand
original (0%)	histogram equalization (29%)	Photoshop auto contrast (0%)	proposed (71%)
keyword = silhouette
original (14%)	histogram equalization (5%)	Photoshop auto contrast (19%)	proposed (62%)

keyword = sand
original (8%)	histogram equalization (3%)	Photoshop auto contrast (8%)	proposed (82%)
keyword = silhouette
original (6%)	histogram equalization (67%)	Photoshop auto contrast (8%)	proposed (19%)

keyword = sand
original (3%)	histogram equalization (43%)	Photoshop auto contrast (3%)	proposed (51%)
keyword = silhouette
original (22%)	histogram equalization (19%)	Photoshop auto contrast (22%)	proposed (36%)

keyword = dark
original (8%)	histogram equalization (0%)	Photoshop auto contrast (3%)	proposed (89%)
keyword = sand
original (14%)	histogram equalization (40%)	Photoshop auto contrast (26%)	proposed (20%)
keyword = silhouette
original (18%)	histogram equalization (5%)	Photoshop auto contrast (26%)	proposed (50%)