Abstract: In accordance with some embodiments of the present invention, de-interlacing may be accomplished by using an edge gradient to select a best interpolation direction for several adjacent pixels between two lines of interlaced video. The interpolation of the intermediate line is performed using the best direction. Then the pixels in the line above and below are analyzed to determine whether the interpolated pixel value should actually be used in view of the possible presence of artifacts.

Abstract: A surveillance system receives video data of a monitored scene and identifies flows of motion in the scene. Specifically, a video camera communicates with a video recognition system, which identifies a flow motion region of a monitored scene and determines characteristics of the identified flow motion region.

Abstract: A color enhancement system may enhance color components in an image. In one embodiment, the color enhancement system suppresses banding by combining an input color value with an enhanced colored value. In another embodiment, local variation in saturation values around a pixel may be used to determine a modification of an input saturation value for color enhancement. As still another embodiment, flickering between frames may be suppressed by combining a histogram that counts the number of times a given pixel value occurs in a particular image with a historical histogram that keeps track of an average of how often given values have occurred in previous frames.

Abstract: A method includes determining a lowest-score interpolation direction among a plurality of interpolation directions. The method further includes calculating a candidate pixel value by interpolating along the lowest-score interpolation direction. The method further includes applying a median function to a set of pixel values. The set of pixel values includes (a) the candidate pixel value, (b) at least one pixel value from a line of pixels that is immediately above a pixel location that is currently being interpolated, and (c) at least one pixel value from a line of pixel values that is immediately below the pixel location that is currently being interpolated.

Abstract: A method includes determining a lowest-score interpolation direction among a plurality of interpolation directions. The method further includes calculating a candidate pixel value by interpolating along the lowest-score interpolation direction. The method further includes applying a median function to a set of pixel values. The set of pixel values includes (a) the candidate pixel value, (b) at least one pixel value from a line of pixels that is immediately above a pixel location that is currently being interpolated, and (c) at least one pixel value from a line of pixel values that is immediately below the pixel location that is currently being interpolated.

Abstract: One implementation of a method for edge directed video de-interlacing in accordance with the disclosed invention includes obtaining at least a portion of a field of input video data including at least portions of four consecutive rows of field pixels including first, second, third, and fourth rows of field pixels. The method further includes selecting an orientation over which to de-interlace the input video data based, at least in part, on a measure of the deviation in pixel values among the four consecutive rows of field pixels and a fifth row of pixels located between the second and third rows of field pixels, the fifth row of pixels including previously interpolated pixel values and pixel values obtained by line averaging between pixel values in the second and third rows of field pixels. The method further includes interpolating along the selected orientation to determine a value for a pixel to be interpolated.

Abstract: The present disclosure describes a method and apparatus for filtering noise in a video image, comprising receiving a first video image, a second video image and motion information wherein the first image may be an image to be filtered and the second image may be a filtered image wherein the second image may be temporally related to said first image.

Abstract: In one embodiment, the present invention includes a method for generating and storing in a lookup table a first array of next field values for use in generation of weighted pixel values and generating and storing a second array in the lookup table of previous field values for use in generation of the weighted pixel values, determining if motion is present in a pixel of a field under analysis, if the motion is present, performing spatial de-noising using a weighted pixel average of a window of a predetermined size, otherwise performing temporal de-noising using a weighted pixel value between the pixel and a previous pixel using information in the lookup table. Other embodiments are described and claimed.

Abstract: In some embodiments, a method includes receiving a plurality of video signal fields, characterizing at least one portion of at least one of the plurality of video signal fields, determining a value for a pixel using inter-field de-interlacing if the characterization satisfies a first criteria, determining a value for a pixel using motion compensated de-interlacing if the characterization satisfies a second criteria, and determining a value for a pixel using intra-field de-interlacing if the characterization satisfies a third criteria. In some embodiments, an apparatus includes a storage medium having stored instructions that when executed by a machine result in the method.

Abstract: Various embodiments for image contrast enhancement are described. In one or more embodiments, adaptive contrast enhancement may be realized by performing local standard deviation (LSD) based histogram equalization. In such embodiments, luminance (Y) values of an image may be grouped into one of several luminance regions, with each luminance region defining an LSD threshold. An LSD value may be calculated for each pixel within a particular luminance region and then compared to the LSD threshold for the particular luminance region. If the calculated LSD value for a pixel is smaller that the defined LSD threshold, then the pixel is not counted as a population for histogram equalization calculation. Otherwise, the pixel is counted as one population. Other embodiments are described and claimed.

Abstract: A system to interpolate between a number of images for the purpose of frame rate conversion may include a motion vector interpolator, a motion compensation module, and a protector. The motion vector interpolator may segment a motion vector into portions based on temporal distances of a target image from each of the number of images. The motion compensation module may generate at least two motion compensated values for a pixel in the target image based on the portions of the motion vector. The protector may choose a motion compensated value for the pixel in the target image when the motion vector is accurate and may choose a pixel value from one of the number of images when the motion vector is not accurate.

Abstract: One implementation of a method for edge directed video de-interlacing in accordance with the disclosed invention includes obtaining at least a portion of a field of input video data including at least portions of four consecutive rows of field pixels including first, second, third, and fourth rows of field pixels. The method further includes selecting an orientation over which to de-interlace the input video data based, at least in part, on a measure of the deviation in pixel values among the four consecutive rows of field pixels and a fifth row of pixels located between the second and third rows of field pixels, the fifth row of pixels including previously interpolated pixel values and pixel values obtained by line averaging between pixel values in the second and third rows of field pixels. The method further includes interpolating along the selected orientation to determine a value for a pixel to be interpolated.

Abstract: The invention is a new method and apparatus that can be used to detect, recognize, and analyze people or other objects in security checkpoints, public-places, parking lots, or in similar environments under surveillance to detect the presence of certain objects of interests (e.g., people), and to identify their activities for security and other purposes in real-time. The system can detect a wide range of activities for different applications. The method detects any new object introduced into a known environment and then classifies the object regions to human body parts or to other non-rigid and rigid objects. By comparing the detected objects with the graphs from a database in the system, the methodology is able to identify object parts and to decide on the presence of the object of interest (human, bag, dog, etc.) in video sequences. The system tracks the movement of different object parts in order to combine them at a later stage to high-level semantics.

Abstract: A method includes determining a lowest-score interpolation direction among a plurality of interpolation directions. The method further includes calculating a candidate pixel value by interpolating along the lowest-score interpolation direction. The method further includes applying a median function to a set of pixel values. The set of pixel values includes (a) the candidate pixel value, (b) at least one pixel value from a line of pixels that is immediately above a pixel location that is currently being interpolated, and (c) at least one pixel value from a line of pixel values that is immediately below the pixel location that is currently being interpolated.

Abstract: The invention is a new method and apparatus that can be used to detect, recognize, and analyze people or other objects in security checkpoints, public-places, parking lots, or in similar environments under surveillance to detect the presence of certain objects of interests (e.g., people), and to identify their activities for security and other purposes in real-time. The system can detect a wide range of activities for different applications. The method detects any new object introduced into a known environment and then classifies the object regions to human body parts or to other non-rigid and rigid objects. By comparing the detected objects with the graphs from a database in the system, the methodology is able to identify object parts and to decide on the presence of the object of interest (human, bag, dog, etc.) in video sequences. The system tracks the movement of different object parts in order to combine them at a later stage to high-level semantics.