Abstract A system for shape recognition includes a spatial convolution filter which has an approximately band pass response. The filter is applied to a camera image and the sign of the filter's output is used to produce a second binary image. At sufficiently coarse (low center spatial frequency) scales, this image tends to change isolated shapes into single blob-shaped regions. The same convolution operator is then reapplied to this binary signal blob image and peaks in the output of the second convolution are used as the primitive elements for building a shape description. By processing the image with operators of two different scales the peaks shift position. The location of peaks from the coarse scale operator together with vectors to the fine scale peaks provide a description of the shape for a look-up table or other system.
A machine method of processing an image containing pixels comprising: applying a first Laplacian of Gaussian convolution to the image to produce a first filtered image; extracting the sign of each pixel in the first filtered image to provide a second filtered image; applying a second Laplacian of Gaussian convolution to the second filtered image to provide a third filtered image; and detecting a first set of peaks in the third filtered image, where said peaks are used to establish a shape description of original image.
Claim 2
Dependent Claim
86 words
A machine method of processing an image containing pixels comprising: applying a first Laplacian of Gaussian convolution to the image to produce a first filtered image; extracting the sign of each pixel in the first filtered image to provide a second filtered image; applying a second Laplacian of Gaussian convolution to the second filtered image to provide a third filtered image; and detecting a first set of peaks in the third filtered image, where said peaks are used to establish a shape description of original image.
Claim 3
Dependent Claim
79 words
A method as in claim 1 further comprising: applying a third Laplacian of Gaussian convolution having a smaller operator to the image to produce a fourth filtered image; extracting the sign of each pixel in the fourth filtered image to provided a fifth filtered image; applying a fourth Laplacian of Gaussian convolution having the same smaller operator to the fifth filtered image to provide a sixth filtered image; detecting a second set of peaks in the sixth filtered image.
Claim 4
Dependent Claim
27 words
A method as in claim 2 further comprising: establishing a description of the image from both the first set of peaks and the second set of peaks.
Claim 5
Dependent Claim
69 words
A method as in claim 3 wherein the step of establishing comprises: detecting which ones of the second set of peaks are disposed within a specified distance of ones of the first set of peaks; and determining the angular relationship between each one of the first set of peaks and those of the second set of peaks disposed within the specified distance of that one of the first set.
Claim 6
Dependent Claim
46 words
A method as in claim 4 further comprising: characterizing the image by determining a best fit line through the first set of peaks, the coordinates of each peak of the first set with respect to a point on the best fit line, and the angular relationship.