System and method for focused-based image segmentation for video signals

An apparatus for generating an output signal representing an outline of pertinent information in a video frame for use in compressing digital signals, said apparatus comprising:

• receiving means for receiving a digital signal, said received digital signal representing a plurality of blocks, each block having an image attribute;
• storage means for storing the image attribute of the blocks;
• and
• processing means for processing said received digital signal to generate said output signal, said processing means operable to:
• store at least a matrix of attributes;
• calculate a focus gradient based, at least in part, on the stored matrix;
• estimate an edge width;
• calculate a focus measurement that is function comprising a ratio of the focus gradient magnitude and the edge width;
• associate with each block a focus value based on the focus measurement;
• compare each focus value with a focus threshold;
• designate each block having a focus value above the focus threshold as focused;
• designate each block having a focus value below the focus threshold as defocused;
• define a focus outline based, at least in part, on the focused blocks.

A prefilter for segmenting a motion video frame into more pertinent and less pertinent portions comprising:

• means for receiving a signal corresponding to an original frame of motion video, the original frame signal having portions, wherein each portion has a depth, and wherein each portion contains an image having an edge characterized by a width and a focus gradient;
• means for measuring the depth of each portion of the original frame signal, wherein the depth measurement is a measure of a degree of focus of the portion, and wherein the degree of focus is determined as a function of edge characteristics, the means for measuring including;
• means for estimating the edge width,
• means for focus gradient magnitude, and
• means for generating a function comprising a ratio of the focus gradient magnitude and the edge width;
• means for designating each portion as foreground or background based, at least in part, on the depth measurement;
• means for generating a template based, at least in part, on the designations;
• means for generating a template signal corresponding to the template;
• and
• means for combining the original frame signal and the template signal into a segmented frame signal.

A method of generating a signal corresponding to a segmented frame comprising:

• acquiring a signal corresponding to an original frame signal comprises an image attribute for each pel;
• comparing the image attribute of each pel to the image attribute of another pel;
• detecting edges in the original signal based on the attribute comparison, wherein the attribute comparison provides an estimate of a focus gradient magnitude and width in which the edge are detected as a function comprising a ratio of the focus gradient magnitude and edge width;
• designating the pels located at edges as focused;
• designating the pels not located at edges as defocused;
• generating a signal corresponding to the focused pels;
• and
• creating a segmented frame signal based on the focused pel signal and the original frame signal.

A method of segmenting an original frame having blocks into a pertinent portion and a less pertinent portion comprising:

• acquiring a signal corresponding to the original frame which signal comprises an image attribute for each block;
• comparing the image attribute of each block to the image attribute of another block;
• assigning each block a focus measurement based at least in part on the attribute comparison, wherein the focus measurement comprises an estimate of edge width;
• comparing the focus measurement of each block to a threshold focus measurement;
• designating the blocks having a focus measurement above the threshold as focused and designating the blocks having a threshold measurement below the threshold as defocused;
• measuring the motion of each block;
• assigning to each block a motion value based on the motion measurement;
• comparing the motion measurement of each block to a threshold motion measurement;
• designating the blocks having a motion measurement above the motion threshold as moving and designating the blocks having a motion measurement below the motion threshold as not moving;
• and
• creating a segmented frame based on the focus and motion designations.

An apparatus for generating an output signal representing an outline of pertinent information in a video frame for use in compressing digital signals, said apparatus comprising:

• receiving means for receiving a digital signal, said received digital signal representing a plurality of blocks, each block having an image attribute;
• storage means for storing the image attribute of the blocks;
• and
• processing means for processing said received digital signal to generate said output signal, said processing means operable to:
• store at least a matrix of attributes;
• calculate a focus gradient based, at least in part, on the stored matrix;
• associate with each block a focus value based on the focus gradient and on estimate of edge width;
• compare each focus value with a focus threshold;
• designate each block having a focus value above the focus threshold as focused;
• designate each block having a focus value below the focus threshold as defocused;
• define a focus outline based, at least in part, on the focused blocks;
• define a focus template by designating the regions between the focused blocks as focused;
• calculate a motion gradient based, at least in part, on the stored matrix;
• associate with each block a motion value based on the motion gradient;
• compare each motion value with a motion threshold;
• designate each block having a motion value above the motion threshold as moving;
• designate each block having a motion value below the threshold as not moving;
• define a motion template based, at least in part, on the moving blocks;
• and
• define a motion/focus outline based on the focus template and the motion template.

A prefilter for segmenting a motion video frame into more pertinent and less pertinent portions comprising:

• means for receiving a signal corresponding to an original frame, the original frame signal having portions, wherein each portion has a depth,
• means for measuring the depth of each portion of the original frame signal based in part on an estimate of edge width;
• means for designating each portion as foreground or background based, at least in part, on the depth measurement;
• means for generating a focus template based, at least in part, on the designations;
• means for generating a focus template signal corresponding to the focus template;
• means for measuring the motion of each portion of the original frame signal;
• means for designating each portion either as moving or nomoving based, at least in part, on the motion measurement;
• means for generating a motion template based, at least in part, on the motion designations;
• means for generating a motion template signal corresponding to the motion template;
• means for comparing the motion template signal and the focus template signal to derive a combined motion/focus template signal;
• and
• means for combining the original frame signal and the combined motion/focus template signal into a combined segmented frame signal.

A method of generating a signal corresponding to a segmented frame comprising:

• acquiring a signal corresponding to an original frame which original frame signal comprises an image attribute for each pel;
• comparing the image attribute of each pel to the image attribute of another pel;
• detecting edges in the original frame signal based on the attribute comparison, wherein the attribute comparison provides an estimate of edge width;
• designating the pels located at the edges as focused;
• designating the pels not located at edges as defocused;
• generating a signal corresponding to the focused pels;
• comparing each pel to a corresponding pel in a different frame;
• measuring the movement of each pel based on the comparison;
• comparing the movement measurement with a movement threshold, designating pels with a movement measurement above the threshold as moving, and designating the pels with a movement measurement below the threshold as nonmoving;
• generating a signal corresponding to the moving pels;
• and
• creating a segmented frame signal based on the focused pel signal, the moving pel signal and the original frame signal.