Pixel value estimation technique using non-linear prediction
AssigneeEastman Kodak Company
InventorDavid R. Cok
Life of this patent
Estimating the term…
Abstract
A pixel value estimation technique uses non-linear prediction to estimate a pixel value before the pixel is known, based on previous values seen. Initially, a number of classes are defined. Each class is a template for a pattern of pixel values. The pixels constituting the support for the prediction of a given pixel are compared to each template. A value giving the goodness of fit to each template is determined as well as a predicted pixel value based on that template. The predictions from each template are averaged together using weights related to the goodness of the fit to that template to produce a consolidated prediction.
Claims8
A method for estimating pixel values in an image, comprising the steps of:
- (a) defining templates for patterns of pixels, each template corresponding to one of the patterns of pixels;
- (b) computing predicted values for a missing pixel in the image for each of the patterns of pixels based on the templates corresponding to the patterns of pixels and on pixel values in a neighborhood of the missing pixel;
- (c) determining a closeness of a match between the templates for each of the patterns of pixels and the pixel values in the neighborhood of the missing pixel;
- (d) producing a pixel value estimation as a finalized estimate of the missing pixel based upon the predicted values and the closeness of the match for each of the templates;
- and
- (e) using the pixel value estimation as the pixel value of the missing pixel in the image.
A method for estimating pixel values in an image, comprising the steps of:
- (a) defining templates for patterns of pixels, each template corresponding to one of the patterns of pixels;
- (b) computing predicted values for a missing pixel in the image for each of the patterns of pixels based on the templates corresponding to the patterns of pixels and on pixel values in a neighborhood of the missing pixel;
- (c) determining a closeness of a match between the templates for each of the patterns of pixels and the pixel values in the neighborhood of the missing pixel;
- (d) producing a pixel value estimation as a finalized estimate of the missing pixel based upon the predicted values and the closeness of the match for each of the templates;
- and
- (e) determining a replacement for the missing pixel value in a compressed image using the pixel value estimation.
Dependencies are inferred from the claim text.