Background recovery in monocular vision

Claim 2

A process for continuous background image recovery in an image sequence of a moving object, said process comprising the steps performed on images of said sequence of: providing a monocular device; inputting said image sequence, via said monocular device, wherein said image sequence contains a moving object moving with respect to a background; identifying at least one region associated with the moving object in said image sequence, wherein said moving object is identified relative to the background; deriving a moving constellation containing said at least one region, and defining said moving constellation in a minimum circumscribing polygon; deriving partial background images associated with respective positions of said moving constellation; and combining said partial background images to create a recovered image corresponding in shape and position to said moving constellation.

Claim 3

A process for background information recovery in accordance with claim 1 wherein said said at least one region of said moving object are identified by comparing differences between successive images in said image sequence.

Claim 4

A process for background information recovery in accordance with claim 2 wherein said partial background images are derived by selecting regions outside of said circumscribing polygon.

Claim 5

A process for background information recovery in accordance with claim 1 wherein said circumscribing polygon is substantially rectangular.

Claim 6

A process for background information recovery in accordance with claim 1 further including the step of deriving the relative depth of said moving object with respect to the background by the first derivative of the size associated with said moving constellation.

Claim 7

A process for background information recovery in accordance with claim 1 further including the step of determining a velocity associated with said moving object by analyzing a perspective compensated image of said moving object.

Claim 8

A process for background image recovery in accordance with claim 1 further comprising the steps of: (1) estimating depth z with the equation 1/(z-kw.sub.i)=l.sub.i /f, wherein T is a threshold value indicative to a minimum number of pixels associated with a display, k is a derived constant for three consecutive equations, l is the length, f is the monocular device focal length and w i is the change in z with respect to a change in time, if 1 i /1 i + 1 ≧T, the threshold value (T), is the number of pixels which a system requires as a minimum difference in pixels; (2) deriving a rough depth estimation and z component velocity estimation using three consecutive frames or every second three frames; (3) compensating perspective distortion for x, y in an image in said image sequence and (4) determining x, y velocity components compared to a previous frame in said image sequence.

Claim 9

A process for background image recovery in accordance with claim 7 further including the step of determining said threshold value as a function of a line size change in frames of said image sequence.

Claim 10

A process for background image recovery in an image sequence of a moving object, said process comprising the steps of: providing a monocular device for providing said image sequence; inputting said image sequence, via said monocular device, wherein said image sequence contains a moving object moving with respect to a background; identifying at least one region associated with the moving object in said image sequence, wherein said moving object is identified from the background; deriving a moving constellation containing said at least one region, and defining said moving constellation in a minimum circumscribing polygon; deriving partial background images associated with respective positions of said moving constellation; and combining said partial background images to derive said background to be recovered; and wherein a background image is recursively obtained and updated in accordance with the algorithm: Mc.sub.i =Rmτ(I.sub.i -I.sub.{i-1}) Bg.sub.0 =I.sub.0 Bg.sub.i= Bg.sub.{i-1} ∪.sub.r (I.sub.i -Mc.sub.i) Bg.sub.unknown =∩.sub.sub{i=0..n} Mc.sub.i where I i denotes input image, Rmτ is a minimum rectangle operator and Mc i is a moving constellation at the ith frame.

Claim 11

A process for background image recovery in an image sequence of a moving object, said process comprising the steps of: providing a monocular device for providing said image sequence; inputting said image sequence, via said monocular device, wherein said image sequence contains a moving object moving with respect to a background; identifying at least one region associated with the moving object in said image sequence, wherein said moving object is identified from the background; deriving a moving constellation containing said at least one region, and defining said moving constellation in a minimum circumscribing polygon; deriving partial background images associated with respective positions of said moving constellation; and combining said partial background images to derive said background to be recovered; and wherein a background image is recursively obtained and updated in accordance with the algorithm: Mc.sub.i =Rmτ(I.sub.i -I.sub.{i-1}) Bg.sub.0 =I.sub.0 Bg.sub.i= Bg.sub.{i-1} ∪.sub.r (I.sub.i -Mc.sub.i) Bg.sub.unknown =∩.sub.sub{i=0..n} Mc.sub.i where I i denotes input image, Rmτ is a minimum rectangle operator and Mc i is a moving constellation at the ith frame.

Claim 12

A process for background image recovery in an image sequence having a plurality of frames, comprising the steps of: providing an optical device for producing said image sequence; inputting said image sequence, via said optical device, wherein said image sequence contains a moving object moving with respect to a background; taking two consecutive frames associated with said image sequence into an image buffer storage continuously at a general video frame rate and storing two consecutive frames; deriving an intensity difference frame from said two consecutive frames; binarizing said intensity difference frame to get a bounding box frame; forming a bounding box in a binarized frame by forming horizontal and vertical projections and thresholding both ends of said bounding box to get boundaries of said bounding box by utilizing a threshold value dependent upon the size of a moving object contained in said image sequence; deriving a partial background frame from a frame associated with said bounding box, wherein regions within the bounding box, except for a region of said bounding box constitute a recovered static background; and overlapping each partial background frame to the background frame to be recovered, thereby incrementally recovering and updating the background every frame.

Claim 13

A process for background image recovery in an image sequence having a plurality of frames, comprising the steps of: providing an optical device for producing said image sequence; inputting said image sequence, via said optical device, wherein said image sequence contains a moving object moving with respect to a background; taking two consecutive frames associated with said image sequence into an image buffer storage continuously at a general video frame rate and storing two consecutive frames; deriving an intensity difference frame from said two consecutive frames; binarizing said intensity difference frame to get a bounding box frame; forming a bounding box in a binarized frame by forming horizontal and vertical projections and thresholding both ends of said bounding box to get boundaries of said bounding box by utilizing a threshold value dependent upon the size of a moving object contained in said image sequence; deriving a partial background frame from a frame associated with said bounding box, wherein regions within the bounding box, except for a region of said bounding box constitute a recovered static background; and overlapping each partial background frame to the background frame to be recovered, thereby incrementally recovering and updating the background every frame.

Claim 14

A process for background image recovery in an image sequence having a plurality of frames, wherein said image sequence includes an image of a moving object, comprising the steps of: providing an optical device for producing said image sequence; inputting said image sequence, via said optical device, wherein said image sequence contains a moving object moving with respect to a background; storing consecutive frames of said plurality of frames including therein said image of the moving object; deriving difference frames from said consecutive frames; deriving a bounding box frame from each difference frame; deriving a partial background frame from each bounding box frame wherein said partial background frame excludes an area within a bounding box; and overlapping said partial background frames for providing said backgrounds information to be recovered.

Claim 15

A process for background image recovery in an image sequence having a plurality of frames, wherein said image sequence includes an image of a moving object, comprising the steps of: providing an optical device for producing said image sequence; inputting said image sequence, via said optical device, wherein said image sequence contains a moving object moving with respect to a background; storing consecutive frames of said plurality of frames including therein said image of the moving object; deriving difference frames from said consecutive frames; deriving a bounding box frame from each difference frame; deriving a partial background frame from each bounding box frame wherein said partial background frame excludes an area within a bounding box; and overlapping said partial background frames for providing said backgrounds information to be recovered.

Claim 16

A process for continuous background image recovery from a sequence of images with moving objects, comprising inputting a sequence of images of a background and a moving object, followed by the steps performed on images of said sequence of: a) identifying regions of moving objects relative to the background; b) deriving a moving constellation containing moving objects in a minimum circumscribing polygon; c) deriving partial background images associated with respective positions of said moving constellation; d) combining a plurality of said partial background images; and e) utilizing background images obtained in step d) for the background of step a).

Claim 17

A process for continuous background image recovery from a sequence of images with moving objects, comprising inputting a sequence of images of a background and a moving object, followed by the steps performed on images of said sequence of: a) identifying regions of moving objects relative to the background; b) deriving a moving constellation containing moving objects in a minimum circumscribing polygon; c) deriving partial background images associated with respective positions of said moving constellation; d) combining a plurality of said partial background images; and e) utilizing background images obtained in step d) for the background of step a).

Claim 18

A process for background image recovery in accordance with claim 12 wherein said regions of moving objects are identified by deriving difference images between an input image with moving objects and a background image obtained by the steps of claim 1.

Claim 19

A process for background image recovery in accordance with claim 13 wherein said partial background images are derived by selecting regions outside of said circumscribing polygon and overlapping onto the previously recovered background image.

Claim 20

A process for background image recovery in accordance with claim 12 wherein said circumscribing polygon is a rectangle.

Claim 21

A process for background image recovery in accordance with claim 12 wherein a depth of said moving object is obtained from the first derivative of size of said image of said moving object and the image of said moving object is obtained from a sequence of input images and the corresponding recovered background images.

Claim 22

A process for background image recovery in accordance with claim 12 wherein a velocity of said moving object is obtained from a perspective compensated image of said object and the image of said object is obtained from a sequence of input images and the corresponding recovered background images.

Claim 23

A process for continuous background image recovery from a sequence of images including an image of a moving object, comprising inputting a sequence of images containing a background and a moving object, followed by the steps performed on said image of: a) identifying regions of moving objects from the background; b) deriving a moving constellation containing moving objects in a minimum circumscribing polygon; c) deriving partial background images associated with respective position of said moving constellation; and d) combining a plurality of said partial background images to derive said background to be recovered and wherein a background image is recursively obtained and updated in accordance with the algorithm: Mc.sub.i =Rmτ(I.sub.i -I.sub.(i-1)) Bg.sub.0 =I.sub.0 Bg.sub.i =Bg.sub.(i-1) ∪.sub.r (I.sub.i -Mc.sub.i) Bg.sub.unknown =∩.sub.sub(i=0..n) Mc.sub.i wherein I i denotes input image and Mc i is a moving constellation the ith frame.

Claim 24

A process for continuous background image recovery from a sequence of images including an image of a moving object, comprising inputting a sequence of images containing a background and a moving object, followed by the steps performed on said image of: a) identifying regions of moving objects from the background; b) deriving a moving constellation containing moving objects in a minimum circumscribing polygon; c) deriving partial background images associated with respective position of said moving constellation; and d) combining a plurality of said partial background images to derive said background to be recovered and wherein a background image is recursively obtained and updated in accordance with the algorithm: Mc.sub.i =Rmτ(I.sub.i -I.sub.(i-1)) Bg.sub.0 =I.sub.0 Bg.sub.i =Bg.sub.(i-1) ∪.sub.r (I.sub.i -Mc.sub.i) Bg.sub.unknown =∩.sub.sub(i=0..n) Mc.sub.i wherein I i denotes input image and Mc i is a moving constellation the ith frame.