Differential motion detection method using background image

Claim 2

A differential motion detection method using background image for deriving an image of a moving object by subtracting a background image from an input image which includes said image of the moving object, said background image being obtained by adding the input image and a previous background image both of which are weighted with variable weighting values, respectively, comprising: a first step of defining a region in which said moving object exists as a moving object existing region, defining a region in which said moving object does not exist as a background region, and defining a region in which a sudden change occurs in said input image as an input image sudden change region; a second step of calculating an absolute value at each pixel in an image produced by subtracting said background image from said input image in order to derive a differential/absolute value image; a third step of calculating a mean value and a variance at a pixel in said differential/absolute value image from the levels of pixels included in a small region on said differential/absolute value image, said small region having said pixel located at the center thereof; a fourth step of distinguishing, from said calculated mean value and variance of said pixel, whether said pixel is included in said moving object existing region, said background region, or said input image sudden change region; and a fifth step of changing a method of updating said background image in accordance with whether said pixel is included in said moving object existing region, said background region, or said input image sudden change region.

Claim 3

A differential motion detection method using background image according to claim 1, wherein said fifth step includes a step of updating said background image, using said input image and said background image, by setting a smaller degree of replacing said background image with said input image when said pixel is a pixel included in said moving object existing region and by setting a larger degree of replacing said background image with said input image when said pixel is a pixel included in said input image sudden change region.

Claim 4

A differential motion detection method using background image according to claim 2, wherein: said fourth step includes a step of counting the number of times said pixel is determined to be a pixel included in said input image sudden change region; and said fifth step includes a step of updating said background image by replacing said background image with said input image or by setting a larger degree of replacing said background image with said input image when said number of times exceeds a predetermined value.

Claim 5

A differential motion detection method using background image for deriving an image of a moving object by subtracting a background image from an input image including said image of the moving object, comprising: a first step of defining a region in which said moving object exists as a moving object existing region, defining a region in which said moving object does not exist as a background region, and defining a region in which a sudden change occurs in said input image as an input image sudden change region; a second step of calculating a differential/absolute value which is an absolute value of each pixel in an image produced by subtracting said background image from said input image in order to derive a differential/absolute value image; a third step of mixing said input image with said background image at a predetermined ratio to preliminarily set integration parameters for updating said background image, said third step preliminarily setting each of said integration parameters to a smaller value as said differential/absolute value is larger when said differential/absolute value of each pixel in said differential/absolute value image is larger than a predetermined threshold value, and preliminarily setting each of said integration parameters to a larger value as said differential/absolute value is smaller when said differential/absolute value of each pixel in said differential/absolute value image is smaller than the predetermined threshold value; a fourth step of calculating a mean value and a variance of a pixel in said differential/absolute value image from the levels of pixels included in a small region on said differential/absolute value image, said small region having said pixel located at the center thereof; a fifth step of distinguishing, from said calculated mean value and variance of said pixel, whether said pixel is included in said moving object existing region, said background region, or said input image sudden change region; and a sixth step of changing a method of updating said background image in accordance with whether said pixel is included in said moving object existing region, said background region, or said input image sudden change region, said sixth step newly setting said preliminarily set integration parameters to update said background image only when said pixel is a pixel included in either of said moving object existing region and said input image sudden change region.

Claim 6

A differential motion detection method using background image for deriving an image of a moving object by subtracting a background image from an input image including said image of the moving object, comprising: a first step of defining a region in which said moving object exists as a moving object existing region, defining a region in which said moving object does not exist as a background region, and defining a region in which a sudden change occurs in said input image as an input image sudden change region; a second step of calculating a differential/absolute value which is an absolute value of each pixel in an image produced by subtracting said background image from said input image in order to derive a differential/absolute value image; a third step of mixing said input image with said background image at a predetermined ratio to preliminarily set integration parameters for updating said background image, said third step preliminarily setting each of said integration parameters to a smaller value as said differential/absolute value is larger when said differential/absolute value of each pixel in said differential/absolute value image is larger than a predetermined threshold value, and preliminarily setting each of said integration parameters to a larger value as said differential/absolute value is smaller when said differential/absolute value of each pixel in said differential/absolute value image is smaller than the predetermined threshold value; a fourth step of calculating a mean value and a variance of a pixel in said differential/absolute value image from the levels of pixels included in a small region on said differential/absolute value image, said small region having said pixel located at the center thereof; a fifth step of distinguishing, from said calculated mean value and variance of said pixel, whether said pixel is included in said moving object existing region, said background region, or said input image sudden change region; and a sixth step of changing a method of updating said background image in accordance with whether said pixel is included in said moving object existing region, said background region, or said input image sudden change region, said sixth step newly setting said preliminarily set integration parameters to update said background image only when said pixel is a pixel included in either of said moving object existing region and said input image sudden change region.

Claim 7

A differential motion detection method using background image according to claim 4, wherein said sixth step includes a step of newly setting said preliminarily set integration parameters such that a degree of replacing said background image with said input image is made smaller when said pixel is a pixel included in said moving image existing region, and newly setting said preliminarily set integration parameters such that the degree of replacing said background image with said input image is made larger when said pixel is a pixel included in said input image sudden change region.