Boundary extraction method and apparatus

1. A boundary extraction method, comprising:

a gradient estimation step of estimating a gradient of each pixel in a captured image;

a gradient adjustment step of adjusting, by enhancing a gradient of a target boundary of an object contained in the captured image and weakening a gradient of a noise boundary, the estimated gradient, so that the adjusted gradient is considered as a current gradient; and

a boundary extraction step of extracting a boundary of the object based on the current gradient,

wherein in the gradient adjustment step, the estimated gradient is weighed with a weight determined based on a gradient difference between two predetermined adjacent regions of each pixel point to enhance the gradient of the target boundary and weaken the gradient of the noise boundary.

2. The method according to claim 1 , wherein the weight TB is determined based on the following formula:

TB= D t /( D b +ε)

wherein Dt=max{f(Nt)}−min{f(Nt)}, Db=max{f(Nb)}−min{f(Nb)}, Nt and Nb represent the two predetermined adjacent regions of each pixel point, Dt represents a difference between a maximum gradient and a minimum gradient in the region Nt, Db represents a difference between a maximum gradient and a minimum gradient in the region Nb, Dt≧Db, and ε is a predetermined constant.

3. The method according to claim 1 , wherein after the boundary extraction step, the method further comprises:

an error boundary determination step of detecting smoothness of the extracted boundary to determine whether the extracted boundary is error;

a gradient correction step of correcting, in a case that the extracted boundary is error, the current gradient, so that the corrected gradient is considered as the current gradient; and

an iteration step of repeating, based on the current gradient, the boundary extraction step, the error boundary determination step and the gradient correction step until a predetermined criterion is satisfied.

4. The method according to claim 3 , wherein the gradient correction step further comprises:

a boundary tendency prediction substep of predicting, starting from a region for which reliability of the extracted boundary is high, tendency of the boundary of the object in a predetermined direction;

a boundary comparison substep of comparing, in the predetermined direction, each point on the extracted boundary with the predicted tendency, starting from a turning point at which the boundary is detected to be not smooth;

an error section determination substep of determining a section from the turning point to a first point on the extracted boundary the difference between which and the predicted tendency is smaller than a predetermined threshold as an error section; and

an elimination substep of setting a gradient of each point within the determined error section to be zero.

5. A boundary extraction apparatus, comprising:

a processor configured to:

estimate a gradient of each pixel in a captured image;

adjust, by enhancing a gradient of a target boundary of an object contained in the captured image and weakening a gradient of a noise boundary, the estimated gradient, so that the adjusted gradient is considered as a current gradient;

extract a boundary of the object based on the current gradient, and

weigh the estimated gradient with a weight determined based on a gradient difference between two predetermined adjacent regions of each pixel point, thus enhancing the gradient of the target boundary and weakening the gradient of the noise boundary.

6. The apparatus according to claim 5 , wherein the weight TB is determined based on the following formula:

TB= D t /( D b +ε)

wherein Dt=max{f(Nt)}−min{f(Nt)}, Db=max{f(Nb)}−min{f(Nb)}, Nt and Nb represent the two predetermined adjacent regions of each pixel point, Dt represents a difference between a maximum gradient and a minimum gradient in the region Nt, Db represents a difference between a maximum gradient and a minimum gradient in the region Nb, Dt≧Db, and ε is a predetermined constant.

7. The apparatus according to claim 5 , wherein the processor is further configured to:

detect smoothness of the extracted boundary to determine whether the extracted boundary is error;

correct, in a case that the extracted boundary is error, the current gradient, so that the corrected gradient is considered as the current gradient; and

repeat, based on the current gradient, processing in the boundary extraction unit, the error boundary determination unit and the gradient correction unit until a predetermined criterion is satisfied.

8. The apparatus according to claim 7 , wherein the processor is further configured to:

predict, starting from a region for which reliability of the extracted boundary is high, tendency of the boundary of the object in a predetermined direction;

compare, in the predetermined direction, each point on the extracted boundary with the predicted tendency, starting from a turning point at which the boundary is detected to be not smooth;

determine a section from the turning point to a first point on the extracted boundary the difference between which and the predicted tendency is smaller than a predetermined threshold as an error section; and

set a gradient of each point within the determined error section to be zero.