Back to results
US 8,526,746 B1Granted
Near lossless data compression method using nonuniform sampling
AssigneeThe United States of America as represented by the Secretary of the Navy
InventorsPathamadai V. Sankar, Leonard Anthony Ferrari
Life of this patent
Estimating the term…
Abstract
Embodiments in accordance with the invention provide methods of near lossless data compression based on nonuniform sampling. Alternate embodiments in accordance with the invention improve the compression factor of lossy compression methods by using nonuniform sampling.
Claims12
A method comprising:
- obtaining an input of an image signal, said image having a plurality of rows of size N;
- for each row of said image:
- nonuniformly sampling said row of said image to generate a plurality of nonuniform samples of said row, each nonuniform sample having an associated sample value and an associated bit representation;
- determining if said row has a constant value P;
- when said row has a constant value P, generating an {RX, RY} value of {0, P}, where P is the constant value;
- when said row does not have a constant value P, applying a 1-D Laplacian transform of S*{−0.1 0.8 −0.1} to each sample value of said row to generate an L value for each said nonuniform sample;
- determining the absolute value of each said L value;
- determining whether each said absolute value of each said L value is greater than a threshold value T;
- determining an L′ value for each said absolute value greater than said threshold T, wherein L′=the absolute value greater than said threshold T;
- determining all L′ values greater than zero;
- from the L′ values greater than zero, selecting the L′ value of a first sample of the row, the L′ value of a last sample of the row, and all the local maxima and minima of L′ of the row;
- generating an {RX, RY} value for each selected L′ value, where RX is a bit representation of the sample associated with said L′ value and RY is a sample value of the sample associated with said L′ value;
- and
- outputting each said {RX, RY} value for said row.
11 more elements
A device comprising:
- a nonuniform sampling module for nonuniformly sampling each row of a received digitized image and generating one or more {RX,RY} values representing said image row data, wherein said generating one or more {RX,RY} values representing said image row data comprises:
- obtaining an input of a image signal, said image having a plurality of rows of size N;
- for each row of said image:
- nonuniformly sampling said row of said image to generate a plurality of nonuniform samples of said row, each nonuniform sample having an associated sample value and an associated bit representation;
- determining if said row has a constant value P;
- when said row has a constant value P, generating an {RX, RY} value of {0, P}, where P is the constant value;
- when said row does not have a constant value P, applying a 1-D Laplacian transform of S*{−0.1 0.8 −0.1} to each sample value of said row to generate an L value for each said nonuniform sample;
- determining the absolute value of each said L value;
- determining whether each said absolute value of each said L value is greater than a threshold value T;
- determining an L′ value for each said absolute value greater than said threshold T, wherein L′=the absolute value greater than said threshold T;
- determining all L′ values greater than zero;
- from the L′ values greater than zero, selecting the L′ value of a first sample of the row, the L′ value of a last sample of the row, and all the local maxima and minima of L′ of the row;
- generating an {RX, RY} value for each selected L′ value, where RX is a bit representation of the sample associated with said L′ value and RY is a sample value of the sample associated with said L′ value;
- and
- outputting each said {RX, RY} value for said row;
- a generalized discrete wavelet transform (GDWT) module for receiving each said one or more {RX, RY} values and for generating and outputting a wavelet transform value;
- and
- a lossless Huffman/LZW/run length coding module for receiving said wavelet transform value and for generating and outputting a nonuniformly sampled compressed data value.
15 more elements
Dependencies are inferred from the claim text.