Video Pro software is a forensic video analysis software product that can be added to any of our systems. This addition includes a more comprehensive set of forensic video processing filters.

With StarWitnesss Video Pro Software, an incredible range of forensic options is at your fingertips. Obscure details can be made visible. Advanced multi-frame, floating point image processing technology and the streaming filter graph architecture allow for video data processing through multiple filters to provide unmatched video or still-frame enhancement.

Standard filters

Deinterlace

Use to remove interlaced motion artifacts visible when displaying interlaced video content, or when capturing still images from interlaced content.

Async Security Demux

Extracts a single camera view from a sequence of multiplexed views.

Image Stabilization

Adjusts frame position to reduce the effects of small camera movements and jiggle.

Frame Median

Reduces drop-out effects or impulse noise effects, often caused by poor tape or recorder quality.

Frame Average

Removes noise in videos where the scene is stationary or stabilized, and image detail must not be lost.

Edge Contrast

The Edge Contrast filter increases contrast near edges in an image. Example uses are enhancing a tattoo to better see its shape, enhancing an emblem or text (car license plate), or enhancing objects in shadows, in over-exposed video, or images with insufficient light.

Deblur

Reduces the effects of blur caused by linear motion, out-of-focus optics, etc.

Sharpen

Enhances edges, textures, and other high-frequency details.

Brightness/Contrast

Controls lightness/darkness and range of contrast.

Zoom

Magnifies a selected area.

The following filter is located under the Display tab, and is normally used at the end of the filter chain to render the video:

Interactive Video Renderer

Outputs a video stream to be shown in a display window.

Advanced filters

Converters

Color to B/W

The Color to B/W filter converts YUV or RGB color video or still images to single channel gray-scale stream. Use this filter to reduce a color image or video to black-and-white

(gray scale). This filter performs a color space transformation based on human brightness perception. Output value for each pixel is equivalent to 'Y' or "Luma" value in YUV color space.

DIB2SSVideo

Use the DIB2SSVideo filter to convert video streams from AVI files or Microsoft Windows-compatible video capture sources to the format used by Signalscape video filters. The DIB2SSVideo filter converts gray-scale, RGB color and YUV 4:2:2 color

video streams to the data format used by Video Pro.

The Signalscape Video data format is designed to represent uncompressed images with any color space or arbitrary number of channels, depths of 8, 16, or 32 bits per channel, and upper left or lower left image origins. YUV image data is stored spatially uncompressed for greater speed and simplicity in filtering operations.

SSVideo2DIB

Use the SSVideo2DIB filter to convert video streams from Signalscape video filter format to Microsoft Windows DIB format used by the Microsoft Video Renderer, compression codecs, or to save to AVI files. The Signalscape Video data format is designed to represent uncompressed images with any color space or arbitrary number of channels, depths of 8, 16, or 32 bits per

channel, and upper left or lower left image origins. Only color spaces understood by Microsoft Windows can be converted by the SSVideo2DIB filter.

YUV to RGB

The YUV to RGB filter converts video from YUV Color Space to RGB after processing under YUV color space by other Signalscape filters. Analog video capture cards and some video codecs output video using a YUV color format, where channels consist of Luma, Chroma Red, and Chroma Blue. Computer displays typically use the

RGB (Red, Green, Blue) color space. The YUV2RGB filter is used to perform this format conversion at any point in the Signalscape filter graph before the video.

File Input/Output

This category includes source and sink filters that import or export image files.

Image file source

The Image File Source filter is used to import image files for processing.. It reads image files stored in standard file formats such as JPEG, TIFF, and Windows BMP. The filter output pin streams the image file contents to other filters as an uncompressed video data frame. The filter opens and decompresses all files using 8 bits per channel then converts them to floating point format if 32-bit output mode is selected.

Image File Writer

The Image File Writer is used to export one or more image files to disk. It writes images to disk using standard file formats such as JPEG, TIFF, and Windows BMP. The input pin of the filter accepts uncompressed frames of video data.

Processing (Advanced)

2D Bandstop

The 2D Bandstop filter is used for removing periodic signals and patterns in an image or video. The filter applies a separable spatial-domain filter to an image to remove a band of horizontal or vertical frequencies. Horizontal and vertical filters can be

applied independently. The 2D Bandstop filter operates in the spatial domain, but uses frequency sampling to determine its filter coefficient values. The number of filter taps determines the number of frequency bins, and thus the precision of the edge frequency locations.

2D Convolution

The 2D Convolution filter is used to apply any desired 2D convolution operation to each video frame. The filter applies a user-specified convolution kernel to each frame of a video sequence. Filter coefficients are imported from a text file. The filter performs a spatial-domain 2-D FIR filtering operation.

2D Derivatives

The 2D Derivatives filter calculates differentials and gradient magnitudes of an image or each video frame. Thirty-two-bit image/video format should be used. This filter applies a derivative kernel or operation to an image. Thirty-two-bit math is suggested. For multi-channel images (color) the derivative information is calculated independently for each channel.

2D Exp

Use the 2D Exp filter to enhance the image’s reflectance properties while suppressing the effect of changes in illumination. This is useful in images with dark shadows and bright areas or images containing areas void of light such as the entrance to a cave. The filter calculates the exponential of 8-bit or 32-bit values. Thirty-two-bit math is suggested. For multi-channel images (color) the exponential is calculated

independently for each channel.

2D FFT Display

The 2D FFT Display filter is used for identifying periodic signals and patterns in an image or video. Frequency identification allows them to be removed using filters such

as the 2D Notch, Fast Fourier Transform (FTT) Zonal Mask, and 2D Bandstop. This filter performs a two-dimensional FFT on an image and outputs the FFT magnitude as an image of the same dimensions and color depth.

2D Highpass

Use the 2D Highpass filter to remove DC components from an image and respond to fast transitions in brightness. It removes low-frequency image components and passes high-frequency components. Vertical and horizontal filtering can be performed

independently. The filter uses a separable FIR implementation.

2D Ln

Use the 2D Ln filter to enhance an images; reflectance properties while suppressing the effect of changes in light. Use 32-bit image/video format with this filter and a 2D Gain filter preceding this filter. The 2D Ln filter calculates the natural logarithm of 8-bit or 32-bit values. Thirty-two-bit math is suggested. For multi-channel images (e.g. color) the natural logarithm is calculated independently for each channel.

2D Lowpass

Use the 2D Lowpass filter to preserve low-frequency image components and to remove noise, interference, high frequency patterns, or sharp transitions. This filter is especially useful before downsampling an image to prevent aliasing effects. Vertical and horizontal filtering can be performed independently. The filter uses a separable FIR implementation.

2D Median

The 2D Median filter removes impulsive noise while preserving sharp transitions in brightness or color between neighboring regions. It sorts the brightness of pixels in the neighborhood surrounding each image pixel and replaces each pixel with the

value of its neighborhood median. For the sort, 32-bit floating-point images are converted to 16-bit fixed-point representation and converted back again which may

cause accuracy loss in applications demanding higher than 16-bit precision.

2D Multiplier

The 2D Multiplier filter multiplies corresponding pixel values from two or more images (or video streams) of the same dimensions and color depth.

2D Notch

The 2D Notch filter is used to remove periodic signals and patterns in an image or video. The filter applies a separable spatial-domain filter to an image to remove a narrow band of horizontal or vertical frequencies. The horizontal and vertical filters may be applied independently. This filter operates in the spatial domain, but uses frequency sampling to determine

its filter coefficient values. The number of filter taps determines the number of frequency bins, and thus the precision of the notch location.

2D Summer

The 2D Summer filter adds multiple images or video streams. It adds corresponding pixel values from two or more images (or video streams) of the same dimensions and color depth. To subtract one input from another, use 32-bit floating-point mode for the data sources, and insert a gain filter with a gain setting on negative one in series with the signal to be subtracted.

Async Deinterlace

Use the Async Deinterlace filter to eliminate interlaced video artifacts in progressive displays or interlaced stills. By exporting results to a video file, you can play the processed file and inspect every field individually. The Async Deinterlace filter outputs two frames for every input frame: one frame for each field. The Async Deinterlace filter deletes the rows from one video field and replaces these pixels with interpolated estimates calculated from pixels in lines above and below

deleted rows (vertical decimation by two followed by vertical linear interpolation.) This process is performed for both fields; two images are output for each input.

Crop/Resize

The Crop/Resize filter changes dimensions of the video image by stretching the image or by applying a new boundary frame to the existing image. This filter is used when the original image dimensions need to be changed. The Crop/Resize filter uses

linear interpolation to resize images.

FFT Zonal Mask

The FFT Zonal mask filter removes unwanted frequencies from the images it processes, and is used to remove periodic signals and patterns in an image or video. It allows removal of diagonal and multi-directional patterns while preserving as much

of the image spectrum as possible. The filter performs a two-dimensional Fast Fourier Transform (FFT) on an image, masks the specified unwanted frequencies, and performs an inverse FFT.

Gaussian Blur

Use the Gaussian Blur filter as a preprocessing step to reduce noise. This preprocessing is used before performing Edge Detection. Applying a Gaussian Blur filter prior to applying an edge detection operation reduces the noise and emphasizes

larger scale features in final results. The filter convolves an image or each frame in an image sequence with a Gaussian

smoothing kernel. To ensure accurate and useable results 32-bit images/video format should be used. For multi-channel images (color) the convolution is performed independently for each channel.

Histogram

Use the Histogram filter to improve the brightness and contrast of images located in both light and dark areas, covering the full intensity range of the video display. It brings object details out of shadows while other well-lit objects remain clear without

saturation. The filter uses 8-bit values internally for calculations. Thirty-two-bit floating-point images are converted to 8-bit fixed-point representation and converted back again

which may cause accuracy loss in applications demanding higher than 8-bit precision.

Mean

Use the Mean filter to reduce noise as a preprocessing step. This filter convolves an image or each frame in an image sequence with a mean kernel of specified size. To ensure accurate and useable results 32-bit images/video format should be used. For multi-channel images (color) the convolution is performed independently for each channel.

Rotate/Flip

The Rotate/Flip filter rotates an image or video stream, or flips it on its horizontal and/or vertical axis. Use this filter to align an image or video stream to the desired view. The filter uses either 8-bit or 32-bit floating point values internally for calculations, depending on the input type. Bilinear interpolation is used to determine output values located between original pixel locations.

Security Demux

The Security Demux filter pulls a single camera view out of a sequence of multiplexed views. Camera images are identified by similarity in content to a reference image to distinguish them from other scenes. The user may adjust the degree of similarity and region of the image used to determine a match. The default

reference image is the first image received during processing.

The filter calculates the image match based on the sum of squared differences between corresponding image pixel values. The sum of squared differences is divided by the sum of squared pixel intensities for the reference image to calculate a ratio.

This ratio is between zero and one, and is compared to the similarity threshold to determine a match.

Tee

The Tee Filter replicates a signal into multiple, identical streams, each with the same data type and number of channels as the original. This can be useful if you want to view the results of processing the same video in different ways, or to view the data with and without processing, as in this example:

Threshold

Use the Threshold filter to convert gray-scale images to pure black-and white images. The filter compares each pixel to threshold value and replaces it with zero (black) if it is below the threshold or full-scale (white) if it is above the threshold. The output image has the same color space and bit depth as the original image. The threshold is performed per-channel. Color images often have different results per channel and result in output images with different combinations of color primaries at 100% strength.