Back to results
US 2010/0074518 A1Published application
Method and apparatus for encoding image
AssigneeKabushiki Kaisha Toshiba
InventorsAkiyuki Tanizawa, Takeshi Chujoh
Life of this patent
Estimating the term…
Abstract
A value of a selection parameter is sequentially set from among a plurality of parameters included in a modeling equation to derive a quantization matrix. A value of each of the parameters is sequentially derived based on a code amount and an encoding distortion obtained from an encoding using a quantization matrix generated from an updated parameter set.
Claims10
An image encoding apparatus comprising:
- an image acquiring unit that acquires an image signal;
- a modeling-parameter setting unit that selects a selection parameter sequentially from among a plurality of parameters included in a modeling equation for deriving a quantization matrix, and sets a value of the selection parameter to a predetermined value;
- a quantization-matrix calculating unit that calculates the quantization matrix by using the modeling equation for each time the selection parameter is set;
- a tentative-encoding processing unit that performs a tentative encoding of the image signal by using the quantization matrix, and derives a quantization parameter representing a code amount, an encoding distortion, and a quantization scale;
- an approximate-expression estimating unit that derives a first approximate expression approximating a relation between the code amount and the selection parameter and a second approximate expression approximating a relation between the encoding distortion and the selection parameter;
- a code-amount-distortion-expression calculating unit that calculates code-amount-distortion relational expression representing a relation between the code amount and the encoding distortion by using the first approximate expression and the second approximate expression;
- a target-code-amount calculating unit that calculates a target code amount or a target encoding distortion by using the code-amount-distortion relational expression and the quantization parameter;
- a modeling-parameter calculating unit that calculates a value of the selection parameter corresponding to the target code amount by using the first approximate expression, or calculates a value of the selection parameter corresponding to the target encoding distortion by using the second approximate expression;
- an optimized-quantization-matrix calculating unit that calculates an optimized quantization matrix for encoding the image signal based on the calculated value of the selection parameter and the modeling equation;
- and
- a final-encoding processing unit that performs a final encoding of the image signal using the optimized quantization matrix;
- wherein the approximate-expression estimating unit derives the first approximate expression and the second approximate expression in such a manner that each of a logarithm of the code amount and a logarithm of the encoding distortion is linearly approximated with respect to the selection parameter.
9 more elements
(canceled)
5-15. (canceled)
An image encoding method comprising:
- acquiring an image signal;
- selecting a selection parameter sequentially from among a plurality of parameters included in a modeling equation for deriving a quantization matrix, and setting a value of the selection parameter to a predetermined value;
- calculating the quantization matrix by using the modeling equation for each time the selection parameter is set;
- performing a tentative encoding of the image signal by using the quantization matrix, and deriving a quantization parameter representing a code amount, an encoding distortion, and a quantization scale;
- deriving a first approximate expression approximating a relation between the code amount and the selection parameter and a second approximate expression approximating a relation between the encoding distortion and the selection parameter;
- calculating code-amount-distortion relational expression representing a relation between the code amount and the encoding distortion by using the first approximate expression and the second approximate expression;
- calculating a target code amount or a target encoding distortion by using the code-amount-distortion relational expression and the quantization parameter;
- calculating a value of the selection parameter corresponding to the target code amount by using the first approximate expression, or calculating a value of the selection parameter corresponding to the target encoding distortion by using the second approximate expression;
- calculating an optimized quantization matrix for encoding the image signal based on the calculated value of the selection parameter and the modeling equation;
- and
- performing a final encoding of the image signal using the optimized quantization matrix;
- wherein the first approximate expression and the second approximate expression are derived in such a manner that each of a logarithm of the code amount and a logarithm of the encoding distortion is linearly approximated with respect to the selection parameter, in the approximate-expression estimating.
9 more elements
(canceled)
20-30. (canceled)
Dependencies are inferred from the claim text.