JPS62136975A - Digital filtering device - Google Patents

Abstract

PURPOSE:To decrease the deterioration in picture quality at animating while phase deviation is eliminated by forming the filter constitution within two fields and forming each coefficient to be point symmetry pairs around an object point in viewing the overlapped two fields as one picture. CONSTITUTION:A field memory 2 stores sampling point data of a preceding field required for a 2-field filter 3 to filter a present field sampling point and fetches a sample of the present field in a shift register 8 sequentially. In reading the field memory 2 and a buffer memory 5, a sample located on a line being one line above of a sample point which is an object of filtering comes to a middle stage of a shift register 7 and a sample below by one line comes to a middle stage of a shift register 6 by synchronization. Thus, samples required for the filtering are stored in the shift registers 6-8, they are fed to prescribed adders 9, 10, a coefficient is multiplied by multipliers 12-14 and an adder 11 sums the result, which is outputted (4). Thus, no phase shift to horizontal and vertical 2-dimension frequencies is given in the picture.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a digital filtering device for filtering sampled television signals.

Conventional technology The digital filtering device by Yano and Yuyama (``Field offset sampling method % type % unevenness 4e, '85/9) is shown in Figure 4, and the digital filtering device by Achiba and Ishikura (``Still image NTSC
"Interfield interpolation sub-Nyquist sampling method for signals",
Television Society Journal VoL, 34,! 6.

1980) is shown in FIG. 6 as a conventional example.

Both examples relate to sub-Nyquist sampling in which the sampling points of a sampled television signal are thinned out to %, and in this application, the three-dimensional spatiotemporal frequency of the image can be low-pass filtered as a result. It is being done.

In the Yano-Yuyama example (Figure 4), when applying filtering to a certain sample point (in this specification, such a sample point is referred to as a target point), the field of this target point and the surrounding Sample points from a total of three fields are used. Furthermore, when viewing this on one screen by overlapping three fields, the coefficients of each sample point have a point-symmetrical relationship with the target point as the center.

On the other hand, in the Achiba-Ishikura example (Figure 6), sample points from two fields are used as a filter, but when viewed overlapping them on one screen, the arrangement of the coefficients of the sample points is not point symmetric. For filtering, sample points on the same line and field as the target point and sample points one line above or below in the front or rear field are used.

Problems to be Solved by the Invention When considering the phase characteristics of a digital filter that filters the two-dimensional horizontal and vertical frequency space of an image, in the example of Achiba and Ishikura, the filter configuration is asymmetric in the two-dimensional space. Therefore, a linear phase cannot be expected. Particularly in this case, since there is no linear phase in the vertical direction, a phase shift may occur depending on the spatial frequency.

On the other hand, the Yano/Yuyama example does not have this problem. However, since it uses coefficients spread over 3 fields, 2
Compared to in-field filters, strong low-pass filtering is applied in the time axis direction, which poses a problem of image quality deterioration when moving images. In particular, a digital filtering device that uses many fields is inevitably strongly influenced in the time-frequency direction, and if this is not desired, it becomes a big problem.

In view of the above, it is an object of the present invention to provide a digital filtering device that has no phase shift and exhibits less deterioration in image quality when moving images.

Means for Solving the Problems In order to overcome the above-mentioned problems, the present invention configures the filter in two fields, and when these two fields are overlapped and viewed as one screen, the object point is centered. This is a digital filtering device in which each coefficient forms a point-symmetric pair.

Effects According to the configuration of the present invention, each coefficient forms a point-symmetrical pair around the target point, so that it has a linear phase with respect to the horizontal and vertical two-dimensional frequencies of the image, and no phase shift occurs. Furthermore, since the filter has a two-field configuration, the amplitude characteristics of the spatio-temporal frequency on the time axis are less influenced by the time-frequency than in a filter that uses three or more fields.

Embodiment A first embodiment of the digital filtering device of the present invention is described below.
As shown in the figure. This is an implementation of the filter shown in FIG. This operation will be explained below.

Input 1 is a sample value series expressed in 8 bits, which is input to field memory 2 and field filter 3.
is input. Field memory 2 holds previous field sample point data necessary for two-field filter 3 to filter the current field sample point. A two-field filter 3 indicated by a broken line sequentially takes sample values of the current field into a shift register 8. Now, the middle register of the shift register 8, which is composed of three stages, is
Holds sample points that have been filtered.

On the other hand, the sample value of the previous field is stored in the field memory 2.
The data is sequentially shifted through the shift register 6, buffer memory 6, and shift register 7. At this time, field memory 2,
In the read operation to the buffer memory 6, in FIG. 2, the sample value one line above the sample point that is the object of the rising wave is placed in the middle stage of the shift register 6, and the sample value one line below is placed in the shift register 6. The synchronization is done so that it is in the middle of the .

Therefore, the shift register 6.7.8 holds the sample values necessary for filtering, which are sent to a predetermined adder 9.10, after which the result is transferred to the multipliers 12, 13 .
'14, the signals are multiplied by coefficients, and these are integrated by the adder 11, resulting in output 4.

In this embodiment, sample values having the same coefficients are added first, and then the coefficients are multiplied, but the reverse is also possible.

Furthermore, multiplication in this embodiment is possible only by shifting the abtt data, and negative number expression is performed by 1 after bit inversion.
This is done using complement representation that adds , and does not require complex multipliers.

FIG. 3 shows an application of the digital filtering device shown in FIG. 1. FIG. 3 shows a basic example of field offset sub-Nyquist encoding (the name is based on the above-mentioned Yano and Yuyama literature) and decoding device that utilizes the correlation between two fields, and 160 inputs are sampled. Frequency 2 f 3
This is a television signal sampled at 17 is a prefilter.

Then, by downsampling 18, this is reduced to the sampling frequency fs. In other words, the sample points are thinned out. For this purpose, it is necessary to limit the band using the prefilter 17 in order to prevent aliasing distortion due to a decrease in the sampling frequency. The digital filtering device shown in FIG. 1 can be utilized as this prefilter. It can also be used as a post-filter with some modification. The operation of the post-filter is to interpolate the thinned out sample points, which changes the coefficient of the target point shown in Figure 2 from 4 to 0, the overall coefficient from % to K, and It can be obtained by setting the point to be interpolated. This example can be realized by the configuration of upsampling 20, field memory 21, and post-filter 22 shown in FIG. Note that 19 is a transmission medium.

The above is an example of a low-pass filter, but the high-pass filter according to the present invention can also be configured and applied in applications such as high-frequency enhancement of images. When using the inter-field correlation of a converted television signal, the coefficients are arranged so as to have a point-symmetrical effect around the target point, so the horizontal and
No phase shift occurs with respect to vertical two-dimensional frequencies. Furthermore, since it is configured to perform this within two fields, it is possible to guarantee equivalent image quality with a simpler device compared to a digital filter that utilizes the correlation of three or more fields.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a digital filtering device according to the present invention, FIG. 2 is an explanatory diagram of a filter realized by this embodiment, and FIG. 3 is a block diagram of an encoding/decoding device according to an application of this embodiment. FIGS. 4 and 6 are explanatory diagrams of conventional examples of digital filtering devices. 1... Sampled television signal input, 2
...Field memory, 3...2 field filter, 6...Buffer 1 memory, 6,
7.8...Shift register, 9,10.11.
...1 adder, 12゜13.14...multiplier. Figure 2 Figure 3 A C; Figure 4 Meeting X

Claims (1)

[Claims] Horizontal, vertical, and temporal 3 of the sampled television signal
A filter that filters dimensional frequencies, and is expressed as a set of coefficients on two fields: the same field as the sample point to be filtered, and the field before or after it.
A digital filtering device characterized in that, when the two fields are viewed as one screen, all the coefficients form point-symmetric pairs around a sample point to be filtered.