Iterated de-noising for image recovery
AssigneeSeiko Epson Corporation
InventorOnur G. Guleryuz
Life of this patent
Estimating the term…
Abstract
An image recovery algorithm that recovers completely lost blocks in an image/video frame using spatial information surrounding these blocks. One application focuses on lost regions of pixels containing textures, edges and other image features that pose problems for other recovery and error concealment algorithms. The algorithm of this invention is based on the iterative application of a generic de-noising algorithm and does not require any complex preconditioning, segmentation, or edge detection steps. Utilizing locally sparse linear transforms and overcomplete de-noising, good PSNR performance is obtained in the recovery of such regions.
Claims30
A method for recovering missing data in a digital signal, comprising the steps of:
- (a) grouping data elements in at least one region in which at least some data is missing into n layers, where n is an integer greater than or equal to 1;
- (b) assigning an initial value to each missing data element in the at least one region and a threshold which is the standard deviation computed from the surrounding of the outer boundary of layer 1;
- and
- (c) for i=1 to n
- (c)(1) evaluating a plurality of orthogonal transforms over layer i
- (c)(2) thresholding selected transform coefficients using the threshold,
- (c)(3) performing an inverse transform operation to obtain intermediate results of layer i, and
- (c)(4) performing an averaging operation followed by a clipping operation on the intermediate results of layer i obtained in (c)(3) to recover layer i.
An apparatus for recovering lost regions in a digital representation, the apparatus comprising one or more components configured to:
- (a) group data elements in at least one region in which at least some data is missing into n layers, where n is an integer greater than or equal to 1;
- (b) assign an initial value to each missing data element in the at least one region and a threshold which is the standard deviation computed from the surroundings of the outer boundary of layer 1;
- and
- (c) for i=1 to n
- (c)(1) evaluate a plurality of orthogonal transforms over layer i
- (c)(2) threshold selected transform coefficients using the threshold,
- (c)(3) perform an inverse transform operation to obtain intermediate results of layer i, and
- (c)(4) perform an averaging operation followed by a clipping operation on the intermediate results of layer i obtained in (c)(3) to recover layer i.
A machine-readable medium having a program of instructions for directing a machine to perform a process of recovering lost regions in a digital representation, the program comprising:
- (a) instructions for grouping data elements in at least one region in which at least some data is missing into n layers, where n is an integer greater than or equal to 1;
- (b) instructions for assigning an initial value to each missing data element in the at least one region and a threshold which is the standard deviation computed from the surroundings of the outer boundary of layer 1;
- and
- (c) instructions for performing the following operations:
- for i=1 to n
- (c)(1) evaluating a plurality of orthogonal transforms over layer i
- (c)(2) thresholding selected transform coefficients using the threshold,
- (c)(3) performing an inverse transform operation to obtain intermediate results of layer i, and
- (c)(4) performing an averaging operation followed by a clipping operation on the intermediate results of layer i obtained in (c)(3) to recover layer i.
Dependencies are inferred from the claim text.