Research ideas

Playing the keys of the Gabor piano. April 7, 2016.

The idea is to tie together two results, one the recent normalized Gabor result from Mark McCourt, reported this year at HVEI, and another a result from experiments I did on shape perception.

Mark showed that many brightness phenomena could be modeled by assuming an array of human Gabor filters whose outputs are normalized. This was so provocative, since we always model their responses according to contrast sensitivity, which is tuned, not normalized.

The other result is the work I did with Dick Voss on Fractals and shape perception. We found that people recognized namable shapes in cloud-like fractal patterns when the fractal dimension of the edge was in the 1.2 to 1.4 range. Fractal dimension is just the slope of the power distribution curve, or for human vision, a monotonically decreasing weighting on spatial frequency response functions.

Rogowitz, B. E., Voss, R., "Shape perception and low-dimension fractal boundary contours," Proceedings of SPIE Vol. 1249, pp. 387-394 (1990).

So it got me thinking. What if the Gabor sensitivity functions are the basic "keys" to the system, but they get weighted in different ways for different tasks.

To test this, we would need a system that allowed us to vary the weighting of an array of Gabor functions, control inputs and measure perceptual effects.

Hearing out visual movement April 7, 2016