Back to results
US 8,027,533 B2Granted
Method of automated image color calibration
AssigneeSTI Medical Systems, LLC
InventorsWenjing Li, Marcelo Soto-Thompson, Ulf Gustafsson, Holger Lange
Life of this patent
Estimating the term…
Abstract
A method of automated image calibration that corrects for non-uniform illumination and calibrates color that is simple, fast, automated, accurate and reliable. A gray balance algorithm is applied to correct for non-uniform illumination and a color calibration algorithm is then applied to calibrate the human subject data. The system has been applied in multiple clinical sites with different instruments.
Claims3
A method of automated image calibration that corrects for non-uniform illumination of a subject and calibrates color, comprising:
- providing an imaging device with a single instrument light source and a single detector, wherein said light source and said detector are uncalibrated;
- supplying a gray target of known ground truth reflectivity and a color target of known ground truth reflectivity;
- collecting raw image data of said subject and calibration image data in RGB color space under substantially identical illumination conditions, wherein said calibration image data comprises an image of said gray target obtained when said detector records native reflectivity from said gray target and an image of said color target obtained when said detector records native reflectivity from said color target, and wherein said image of a color target contains color patches, each of said patches having a native position value and a standard color value for said RGB color space;
- applying a gray balance algorithm which uses said image of said gray target to said raw image data and said image of a color target to produce a gray balanced raw image and a gray balanced image of a color target, wherein said gray balanced raw image and said gray balanced image of a color target are flat-field images with normalized light distribution;
- and
- performing color transformation by extracting said native position value and determining a native color value for each of said color patches, calculating a color correction matrix between said native color values and said standard color values, and applying said matrix to said gray balanced raw image to correct for color inaccuracies, without knowledge of characteristics of said light source and said detector.
3 more elements
A method of automated image calibration that corrects for non-uniform illumination of a subject and calibrates color, comprising:
- providing an imaging device with a single instrument light source and a single detector, wherein said light source and said detector are uncalibrated;
- supplying a gray target of known ground truth reflectivity and a color target of known ground truth reflectivity;
- collecting raw image data of said subject and calibration image data in RGB color space under substantially identical illumination conditions, wherein said calibration image data comprises an image of said gray target obtained when said detector records native reflectivity from said gray target and an image of said color target obtained when said detector records native reflectivity from said color target;
- converting said raw image data and said calibration image data to a second color space, wherein said image of a color target contains color patches, each of said patches having a native position value and a standard color value for said second color space;
- applying a gray balance algorithm which uses said image of said gray target to said raw image data and to said image of a color target to produce a gray balanced raw image and a gray balanced image of a color target, wherein said gray balanced raw image and said gray balanced image of a color target are flat-field images with normalized light distribution;
- and
- performing color transformation by extracting said native position value and determining a native color value for each of said color patches, calculating a color correction matrix between said native color values and said standard color values, and applying said matrix to said gray balanced image raw image data to correct for color inaccuracies, without knowledge of characteristics of said light source and said detector.
4 more elements
Dependencies are inferred from the claim text.