Limits...
An exponential filter model predicts lightness illusions.

Zeman A, Brooks KR, Ghebreab S - Front Hum Neurosci (2015)

Bottom Line: We investigate whether predictive success depends on filters of a particular size or shape and whether pooling information across filters can improve performance.Combining two filters together increased the best performance to 23, with asymptotic performance at 24 for an arbitrarily large combination of filter outputs.While normalization improved prediction magnitudes, it only slightly improved overall scores in direction predictions.

View Article: PubMed Central - PubMed

Affiliation: Department of Cognitive Science, ARC Centre of Excellence in Cognition and its Disorders, Macquarie University Sydney, NSW, Australia ; Commonwealth Scientific and Industrial Research Organisation Marsfield, NSW, Australia ; Perception in Action Research Centre, Macquarie University Sydney, NSW, Australia.

ABSTRACT
Lightness, or perceived reflectance of a surface, is influenced by surrounding context. This is demonstrated by the Simultaneous Contrast Illusion (SCI), where a gray patch is perceived lighter against a black background and vice versa. Conversely, assimilation is where the lightness of the target patch moves toward that of the bounding areas and can be demonstrated in White's effect. Blakeslee and McCourt (1999) introduced an oriented difference-of-Gaussian (ODOG) model that is able to account for both contrast and assimilation in a number of lightness illusions and that has been subsequently improved using localized normalization techniques. We introduce a model inspired by image statistics that is based on a family of exponential filters, with kernels spanning across multiple sizes and shapes. We include an optional second stage of normalization based on contrast gain control. Our model was tested on a well-known set of lightness illusions that have previously been used to evaluate ODOG and its variants, and model lightness values were compared with typical human data. We investigate whether predictive success depends on filters of a particular size or shape and whether pooling information across filters can improve performance. The best single filter correctly predicted the direction of lightness effects for 21 out of 27 illusions. Combining two filters together increased the best performance to 23, with asymptotic performance at 24 for an arbitrarily large combination of filter outputs. While normalization improved prediction magnitudes, it only slightly improved overall scores in direction predictions. The prediction performance of 24 out of 27 illusions equals that of the best performing ODOG variant, with greater parsimony. Our model shows that V1-style orientation-selectivity is not necessary to account for lightness illusions and that a low-level model based on image statistics is able to account for a wide range of both contrast and assimilation effects.

No MeSH data available.


Related in: MedlinePlus

Exponential filters applied to White's illusion, all with sizeK2 = 5. The top row shows a filter with high kurtosis (m = 0.5), the middle row shows a medium kurtosis filter (m = 1.0) and the bottom row shows a low kurtosis filter (m = 2.0). From left to right, column 1 is a top-down view of the filter shape, column 2 is the original image (of size 512 × 512 pixels), column 3 is the same image filtered and column 4 is a cross section of grayscale values through row y = 250 pixels (where 0 represents black and 255 represents white). The locations of target patches are highlighted yellow in the final column.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4478851&req=5

Figure 4: Exponential filters applied to White's illusion, all with sizeK2 = 5. The top row shows a filter with high kurtosis (m = 0.5), the middle row shows a medium kurtosis filter (m = 1.0) and the bottom row shows a low kurtosis filter (m = 2.0). From left to right, column 1 is a top-down view of the filter shape, column 2 is the original image (of size 512 × 512 pixels), column 3 is the same image filtered and column 4 is a cross section of grayscale values through row y = 250 pixels (where 0 represents black and 255 represents white). The locations of target patches are highlighted yellow in the final column.

Mentions: Responses are then convolved to create a filtered image of the same dimensions as the original input. The filtered convolved image is subtracted from the original image as the final step in processing. We explore a range of different filter shapes and sizes and produce a set of filtered images for every size and shape of filter. We use 10 filter sizes ranging from 5 pixels to 95 in increments of 10. The filter shapes range from 0.1 to 1.9 in increments of 0.2. Figure 4 illustrates the result of applying three example filters with different shape parameters to White's Illusion. The predictive success of this particular filter size is well-demonstrated for this particular image, regardless of filter shape. The bottom row in Figure 4 demonstrates a close approximation to the Gaussian filter, which in this case is able to predict the direction and magnitude of White's Effect. This filter differs from the DOG filters used by Blakeslee and McCourt (1997)'s model in two key ways. Firstly, Blakeslee and McCourt use a Difference-of-Gaussian (DOG) filter, rather than an approximate Gaussian pictured here. Secondly, Figure 4 demonstrates a single filter operation, rather than a bank of filters used by Blakeslee and McCourt (1997).


An exponential filter model predicts lightness illusions.

Zeman A, Brooks KR, Ghebreab S - Front Hum Neurosci (2015)

Exponential filters applied to White's illusion, all with sizeK2 = 5. The top row shows a filter with high kurtosis (m = 0.5), the middle row shows a medium kurtosis filter (m = 1.0) and the bottom row shows a low kurtosis filter (m = 2.0). From left to right, column 1 is a top-down view of the filter shape, column 2 is the original image (of size 512 × 512 pixels), column 3 is the same image filtered and column 4 is a cross section of grayscale values through row y = 250 pixels (where 0 represents black and 255 represents white). The locations of target patches are highlighted yellow in the final column.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4478851&req=5

Figure 4: Exponential filters applied to White's illusion, all with sizeK2 = 5. The top row shows a filter with high kurtosis (m = 0.5), the middle row shows a medium kurtosis filter (m = 1.0) and the bottom row shows a low kurtosis filter (m = 2.0). From left to right, column 1 is a top-down view of the filter shape, column 2 is the original image (of size 512 × 512 pixels), column 3 is the same image filtered and column 4 is a cross section of grayscale values through row y = 250 pixels (where 0 represents black and 255 represents white). The locations of target patches are highlighted yellow in the final column.
Mentions: Responses are then convolved to create a filtered image of the same dimensions as the original input. The filtered convolved image is subtracted from the original image as the final step in processing. We explore a range of different filter shapes and sizes and produce a set of filtered images for every size and shape of filter. We use 10 filter sizes ranging from 5 pixels to 95 in increments of 10. The filter shapes range from 0.1 to 1.9 in increments of 0.2. Figure 4 illustrates the result of applying three example filters with different shape parameters to White's Illusion. The predictive success of this particular filter size is well-demonstrated for this particular image, regardless of filter shape. The bottom row in Figure 4 demonstrates a close approximation to the Gaussian filter, which in this case is able to predict the direction and magnitude of White's Effect. This filter differs from the DOG filters used by Blakeslee and McCourt (1997)'s model in two key ways. Firstly, Blakeslee and McCourt use a Difference-of-Gaussian (DOG) filter, rather than an approximate Gaussian pictured here. Secondly, Figure 4 demonstrates a single filter operation, rather than a bank of filters used by Blakeslee and McCourt (1997).

Bottom Line: We investigate whether predictive success depends on filters of a particular size or shape and whether pooling information across filters can improve performance.Combining two filters together increased the best performance to 23, with asymptotic performance at 24 for an arbitrarily large combination of filter outputs.While normalization improved prediction magnitudes, it only slightly improved overall scores in direction predictions.

View Article: PubMed Central - PubMed

Affiliation: Department of Cognitive Science, ARC Centre of Excellence in Cognition and its Disorders, Macquarie University Sydney, NSW, Australia ; Commonwealth Scientific and Industrial Research Organisation Marsfield, NSW, Australia ; Perception in Action Research Centre, Macquarie University Sydney, NSW, Australia.

ABSTRACT
Lightness, or perceived reflectance of a surface, is influenced by surrounding context. This is demonstrated by the Simultaneous Contrast Illusion (SCI), where a gray patch is perceived lighter against a black background and vice versa. Conversely, assimilation is where the lightness of the target patch moves toward that of the bounding areas and can be demonstrated in White's effect. Blakeslee and McCourt (1999) introduced an oriented difference-of-Gaussian (ODOG) model that is able to account for both contrast and assimilation in a number of lightness illusions and that has been subsequently improved using localized normalization techniques. We introduce a model inspired by image statistics that is based on a family of exponential filters, with kernels spanning across multiple sizes and shapes. We include an optional second stage of normalization based on contrast gain control. Our model was tested on a well-known set of lightness illusions that have previously been used to evaluate ODOG and its variants, and model lightness values were compared with typical human data. We investigate whether predictive success depends on filters of a particular size or shape and whether pooling information across filters can improve performance. The best single filter correctly predicted the direction of lightness effects for 21 out of 27 illusions. Combining two filters together increased the best performance to 23, with asymptotic performance at 24 for an arbitrarily large combination of filter outputs. While normalization improved prediction magnitudes, it only slightly improved overall scores in direction predictions. The prediction performance of 24 out of 27 illusions equals that of the best performing ODOG variant, with greater parsimony. Our model shows that V1-style orientation-selectivity is not necessary to account for lightness illusions and that a low-level model based on image statistics is able to account for a wide range of both contrast and assimilation effects.

No MeSH data available.


Related in: MedlinePlus