# US 2012/0251003 A1 · Image processing system and method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2012/0251003 A1
- Document ID: US20120251003A1
- Status: Published application
- Kind: A1
- Application number: 13/408,541
- Filed: 2012-02-29
- Published: 2012-10-04
- Earliest claimed: 2011-03-23
- Assignees: Kabushiki Kaisha Toshiba
- Inventors: Frank Perbet; Atsuto Maki; Minh-Tri Pham; Bjorn Stenger; Oliver Woodford
- Web page: https://patentgrep.com/patent/US20120251003A1

## CPC

- G06T 7/11: Region-based segmentation
- G06F 18/2323: based on graph theory, e.g. minimum spanning trees [MST] or graph cuts
- G06T 7/162: involving graph-based methods
- G06T 7/187: involving region growing; involving region merging; involving connected component labelling
- G06V 10/7635: {based on graphs, e.g. graph cuts or spectral clustering}
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

A method dividing an image into plural superpixels of plural pixels of the image. The method calculates an initial set of weights from a measure of similarity between pairs of pixels, from which a resultant set of weights is calculated for pairs of pixels that are less that a threshold distance apart on the image. The calculation calculates a weight for a pair of pixels as the sum over a set of third pixels of the product of initial weight of the first pixel of the pair of pixel with the third pixel and the weight of the third pixel with the second pixel. Each weight is then subjected to a power coefficient operation. The resultant set of weights and the initial set of weights are then compared to check for convergence. If the weights converge, the converged set of weights is used to divide the image into superpixels.

## Claims

### Claim 1

A method of dividing an image into a plurality of superpixels, each superpixel comprising a plurality of pixels of the image, the method comprising
calculating an initial set of weights, each weight being associated with a pair of pixels of the image, using a measure of similarity between neighbouring pixels;
calculating from the initial set of weights a resultant set of weights for pairs of pixels in the image separated by less than a threshold distance, wherein calculating the resultant set of weights comprises
calculating a weight for a first pixel and a second pixel as the sum over third pixels of the product of the initial weight between the first pixel and a third pixel with the initial weight between that third pixel and the second pixel;
calculating a value of each weight to a power coefficient; and
normalising the set of weights to the power coefficient such that the set of all weights of a pixel sum to one;
comparing the initial set of weights and the resultant set of weights to check for convergence to a converged set of weights; and
grouping the pixels into superpixels using the converged set of weights.

### Claim 2 (depends on 1)

A method according to claim 1, wherein calculating a weight for a first pixel and a second pixel comprises reading a first look up table, the first look up table indicating a set of transformations on the image from the first pixel to the set of pixels that are within the threshold distance of the first pixel.

### Claim 3 (depends on 2)

A method according to claim 2, wherein calculating a weight for a first pixel and a second pixel further comprises reading a second look up table, the second look up table indicating the set transformations on the image from the first pixel to the second pixel via the set of third pixels.

### Claim 4 (depends on 1)

A method according to claim 1, wherein the measure of similarity between pixels is calculated from a difference in intensity between pixels.

### Claim 5 (depends on 1)

A method according to claim 1, further comprising allocating memory for storing a set of weights associated with each pixel of the image, the memory being allocated according to the number of pixels with the threshold distance of a given pixel.

### Claim 6 (depends on 5)

A method according to claim 5, further comprising storing the initial set of weights in the memory as the set of weights associated with each pixel and replacing the initial set of weights with the resultant set of weights following calculation of the resultant set of weights.

### Claim 7 (depends on 1)

A method of processing an image comprising
dividing image into superpixels using a method according to claim 1; and
processing the image using the superpixels.

### Claim 8 (depends on 7)

A method according to claim 7, wherein processing the image using the superpixels comprise recognising a feature in the image.

### Claim 9 (depends on 7)

A method according to claim 7, wherein processing the image using the superpixels comprises encoding the image.

### Claim 10 (depends on 7)

A method according to claim 7, wherein processing the image using the superpixels comprises calculating a three dimensional reconstruction of the image.

### Claim 11 (depends on 7)

A method according to claim 7, wherein processing the image using the superpixels comprises reducing noise in the image.

### Claim 12 (depends on 1)

A computer readable medium carrying computer executable instructions which when executed on a computer cause the computer to execute a method according to claim 1.

### Claim 13

An image processing system for dividing image into a plurality of superpixels, each superpixel comprising a plurality of pixels of the image, the system comprising:
a processor operable to
calculate an initial set of weights, each weight being associated with a pair of pixels of the image, using a measure of similarity between neighbouring pixels;
calculate from the initial set of weights a resultant set of weights for pairs of pixels in the image separated by less than a threshold distance, by calculating a weight for a first pixel and a second pixel as the sum over third pixels of the product of the initial weight between the first pixel and a third pixel with the initial weight between that third pixel and the second pixel; calculating a value of each weight to a power coefficient; and normalising the set of weights to the power coefficient such that the set of all weights of a pixel sum to one;
compare the initial set of weights and the resultant set of weights to check for convergence to a converged set of weights; and
group the pixels into superpixels using the converged set of weights.

### Claim 14 (depends on 13)

A system according to claim 13, further comprising storage for a first look up table, the first look up table indicating a set of transformations on the image from the first pixel to the set of pixels that are within the threshold distance of the first pixel, the processor being operable to calculate a weight for a first pixel and a second pixel by reading a first look up table.

### Claim 15 (depends on 14)

A system according to claim 14, further comprising storage for a second look up table, the second look up table indicating the set transformations on the image from the first pixel to the second pixel via the set of third pixels, the processor being operable to calculate a weight for a first pixel and a second pixel by reading the second look up table.

### Claim 16 (depends on 13)

A system according to claim 13, further comprising a memory, the memory being allocated for storing a set of weights associated with each pixel of the image being allocated according to the number of pixels with the threshold distance of a given pixel.

### Claim 17 (depends on 16)

A system according to claim 16, the processor being further operable to store the initial set of weights in the memory as the set of weights associated with each pixel and replace the initial set of weights with the resultant set of weights following calculation of the resultant set of weights.

### Claim 18 (depends on 13)

A system according to claim 13, the processor being further operable to reduce noise in the image using the superpixels.

### Claim 19 (depends on 13)

A system according to claim 13, wherein the processor is operable to recognise a feature in the image using the superpixels.

### Claim 20 (depends on 13)

A system according to claim 13, wherein the processor is operable to calculate a three dimensional representation of the image using the superpixels.
