FreqFilter plugin

Plugin BPP_FreqFilter – the most common type of frequency filter designed to remove periodic noise. Periodic noise occurs when harmonic oscillation of different frequencies and phases applied to the image, and looks on the screen as moir, ripples or diagonal transparent slow moving or stationary strips. The cause of such interference is a mismatch in the distribution of cable networks, the installation of low-quality antenna splitters and amplifiers.

Due to the relatively slow-changing pattern of periodic noise temporal noise suppressors mistakenly consider such interference as a part of the image and do not have a serious influence on them. To remove periodic interference by shumodava space, they must be sufficiently large radius values and other inherent in this class of filter parameters, which inevitably leads to the destruction of the image.

To suppress periodic noise frequency filters are used, which work with the image in the frequency domain.

To understand how the frequency filter works, we must clear about that any bar, any point on the screen can be represented by a set of frequencies – sinusoidal oscillations with a certain amplitude and the initial phase. Lower frequencies are involved in the formation of large fragments of images, higher – small. The set of frequencies involved in the formation of the image, called the spectrum of the signal, the individual frequency – harmonics. The more high frequency harmonics in the spectrum, the sharper the image and it has more fine detail. The amplitudes of the harmonics, from which an image is formed along the axes X and Y, can be seen in the plugin BPP_FreqFilter.

In contrast to the dynamically changing the image of the spectrum of the main signal, a harmonic of the periodic interference is constant in time and frequency, and usually stands out against the background of the overall spectrum. If interference is strictly a constant frequency and phase, then it looks on the screen in the form of fixed or floating slowly transparent vertical strips. On the spectrum diagram this noise appears as a bright two fixed points. Tilt of strips on the screen determines the location of the points on the Y axis in range.

If the interference is frequency and phase modulated signal, for example, traces of sound or color subcarrier of image, the point on the spectrum is blurred in a vertical line or a strip. On the screen, this noise appears as moire.

In order to remove periodic interference, a television signal converted to the frequency domain, harmonic interference is localized and suppressed in amplitude. Then the spectrum is converted back to a signal of the usual kind. This procedure is called filtering in the frequency domain. Rejection ratio or gain of any region of the spectrum or the harmonic set by filter. The plugin allows you to build three types of elementary filters:

«Dot» – spot Butterworth filter designed to suppress the individual harmonics. Filter parameters are: center frequency of the filter window «Fx» and «Fy», the width of the filter window horizontally and vertically «Wx» and «Wy», the filter order «Order» and depth filtration «Strength». Filter «Dot» has an additional feature. When the «Strength» value is equal to zero, the Butterworth window is replaced by adaptive function which is not cut, but masked selected harmonic, based on the average surrounding its spectrum.

«Win 1D» – bandpass Butterworth filter designed to suppress the periodic noise with variable frequency and phase. This filter works only with lines and has no effect on the image along the vertical axis. The filter is characterized by a center frequency of the filter window «Fx» (parameter «Fy» ignored), the width of the filter window «Wx» (parameter «Wy» ignored), order and depth of filtration.

«Win 2D» – bandpass Butterworth filter for amplifying / suppression frequency band along the X and Y axes of the image. The filter can be used to improve sharpness or image smoothing in its horizontal and vertical planes. Filter parameters are: the location and bandwidth of the filter along the axis X, the position and width of Y, order and depth filtration.

The plugin allows you to create an unlimited number of elementary filters and thereby make the total filter of any desired shape. To create a simple filter, click on the button "Add filter" and choose in the future position of the spectrum window. Default created point filter type «Dot», which subsequently can be converted to any other type. When choosing an elementary filter its parameters are transferred to the editing line up on the list. Enable / disable the filter and select the desired parameter for editing is possible using a combination of hot keys «Ctrl» + «` »...« 8 ».

To convert the signal to the frequency domain using fast Fourier transform algorithm (FFT). The algorithm processes the image by lines whose length is fixed and equal to a power of two – i.e. 256, 512 and 1024. The longer the line, the larger the picture can be processed without loss of quality. Losses arise if the horizontal size of the frame is larger than the length of the FFT line. CPU load is proportional to the length of the line. When viewing live channels in order to optimize CPU load line length can be reduced to 512 without noticeable loss of quality

About of image spectrum.

As noted above, the quality and sharpness of the image depends on the width and the filling of its frequency spectrum. Digitizing video processed by tuner at a frequency 13.5MHz. One TVL has a duration of 52 ms or 702 reference digitized signal. Accordingly, the maximum spectrum width of the digitized television signal will be equal 13.5 / 2 = 6.75MHz. Luma component of the television image occupies only part of the television signal spectrum. At frequencies above 4MHz passed information about the color and sound of TV programs. When restoring the TV signal is removed from it all the "extra" portion of the spectrum and then in the above 3.5MGts SECAM system typically does not contain any useful information.

At frequencies 4.25MHz and 4.406MHz clearly visible traces of not completely suppressed color subcarrier of television signal. Remains of subcarriers cause ripples in the areas of the image with rich color and sharp transitions. The greatest disadvantage of this is shown in the TV system SECAM. Owners pay TV tuners built on chipsets SAA713x, available hardware "Antialias filter for SECAM», which solves the problems with sub-carriers, however, due to a minor loss of image sharpness.

You can create your antialias filter based on «Win 1D» filter of the plugin.

Information about the spectrum of the video can be useful when recording. Knowing her, it is possible to some extent to reduce the size of the frame and thus save on bitrate and therefore on the size of the target file. The screenshots below show the spectral characteristics of the same frame VHS recording for different video dimensions – 704x576, 480x576 and 352x576, respectively.

Despite the fact that reducing the frame size to 480x576 led to a narrowing of the spectrum, the picture quality for that resolution has not changed since 4.6MHz band is more than enough to transfer all the frequency characteristics of the video standard VHS. In this case, the frame size of 704x576, with its maximum spectral width 6.75MHz manifestly excessive. If desired, the frame size can be reduced even more, up to 352x576. At the same time outside the bandwidth 3.4MHz are remnants of the color subcarrier PAL system in the area 4.5MHz. On the one hand this is good, on the other hand, at this resolution is already begins to affect the spectrum of the useful signal, which becomes noticeable to the eye.

Thus, without substantial damage to the quality of the image while recording a frame size can be reduced to a resolution of 480 pixels horizontally. That resolution 480x576 accepted as the standard SVCD how best to digitize terrestrial SECAM television signal and VHS recordings.

If we consider the SD quality TV signal of the PAL system, taken from the video output of the DVD player, satellite receiver, game console or DVB set-top box, it will take almost the entire allotted for his band 6.0-6.5MHz. In this case, save on reducing of the frame size does not turn.

Complete removal or change of the harmonic's amplitude does not pass completely and gives rise to distortion in the form of the so-called "ringing" – false contours on the borders of abrupt transitions of contrasting objects on the screen. The magnitude of ringing depends on the size of the filter window, its proximity to the origin of the spectrum and filter order. The larger the window, the smoother filter function and less distortion. However, a large window together with interference captures and suppresses neighboring frequencies that leads to blurring of the image. During set up filters are often not necessary to complete suppression of any frequency. If target harmonic has a constant amplitude, so that it is enough to choose the option «Strength» filter to interference just did not stand out in the background surrounding it. In the case of spot interference ideal option is to use an adaptive filter, which shows good results if an noise harmonic is located in the inactive region of the spectrum.

List of changes

• First release.

• Added filter visualization.

Link to the plugin: BPP_FreqFilter_v101.zip (61.9 kb)

24.10.14