JPS6337708A - Band limitation interpolating filter - Google Patents

Abstract

PURPOSE:To shorten the calculation time for finding a tap coefficient and to vary band characteristics of a filter only varying intervals of table reference by employing a method for referring tables which contain impulse responses. CONSTITUTION:An image memory 5 outputs 2I'+1 image data Pi' based upon a specified address io as their center. The image data Pi' outputted by the image memory 5 are delayed by an FIFO 12 by the time required to find the coefficient of a filter. The tables T-1'-TI ' store discrete data which are obtained by dividing the data into Tx' up to the 1st zero point of a Sinc function and sampling them with 1/2Tx'. Tap coefficients of filters obtained by referring to the tables and said image data Pi' are multiplied by multiplexers Mo', and M-I-MI ' respectively and their sum is calculated by an adder Ao to obtain an output image qmo, which is outputted to an output terminal 2.

Description

Detailed Description of the Invention [Industrial Field of Application] The present invention relates to a method for band limiting and interpolation used in conjunction with a geometric conversion device for performing geometric conversion processing of television images at high speed and with high image quality. It is related to filters.

[Prior Art] Conventionally, this type of geometrical transformation device has been devised as a special effects device for television signals such as a digital video effector (DVE).

FIG. 2 is a block diagram showing an example of a conventional video signal enlargement/reduction device of this type.

In the figure, 1 is an input terminal for a television video signal; 2 is an input terminal for a television video signal;
3 is an output terminal for a television video signal, 3 is an A/D converter, and 4 is a band-limiting spatial filter, which reduces aliasing noise caused by sampling. 5 is image memory, 6v
and 6H are vertical and horizontal interpolation filters, which respectively reduce noise due to interpolation processing. 9.10 and 11 are switches for changing the enlargement/reduction of the image.

The operation of this conventional example will be described below. That is, when enlarging, images written in the image (frame) memory 5 are read out at intervals smaller than the intervals at which they were written. At this time, new pixel data is obtained through interpolation processing by the vertical interpolation filter 8V and the horizontal interpolation filter 6H.

When reducing, pixels are thinned out after band-limiting processing is performed by the band-limiting spatial filter 4 to prevent aliasing.
New pixel data is obtained through interpolation processing by a vertical interpolation filter 6V and a horizontal interpolation filter δH, and is written into the frame memory 5.

Recently, memory access times have become faster, and a mapping method using pixel positions determined by coordinate transformation as image memory addresses has become possible.

As an example of such an apparatus, FIG. 3 is a block diagram showing the configuration of another example of geometric conversion processing of a video signal.

In the figure, the same parts as in FIG. 2 are given the same reference numerals, and the explanation thereof will be omitted. 6 is a spatial interpolation filter, 7 is D/
The A converter 8 is an address generator that generates an address for reading out the image memory 5 in order to geometrically transform the image.The operation of this other conventional example will be described below.

The manually generated image is band-limited by a band-limiting spatial filter 4 to prevent aliasing, and then the image memory 5
will be written to. In the geometric transformation processing, the integer part of the new pixel coordinates obtained by coordinate transformation is used as the address of the image memory 5, and the pixel data is read out by the address generator 8, and the information of the decimal point of the coordinates is used to read the pixel data from the spatial interpolation filter 6.
Perform interpolation by

In the above device, the band limiting process by the band limiting spatial filter 4 uses a sine square filter of about 3rd to 7th order to simplify the circuit configuration. Furthermore, in order to adapt the characteristics to the reduction rate, several filters with different band limit frequencies are switched. This complicates the configuration of the adaptive filter and increases the cost, so the circuit configuration in which the sine square filter is switched in this way is used to simplify the configuration. The interpolation processing by the spatial interpolation filter 6 is performed at the level of linear interpolation processing, thereby simplifying the circuit configuration.

Since the band-limiting filter 4 and the interpolation filter 6 have different purposes, they have circuit configurations that perform different processes, and in the conventional filter configuration, they are independent from each other.

[Problems to be Solved by the Invention] There are two problems with conventional devices that perform geometric transformation of images in real time.

First, during the reduction process, filter characteristics are pseudo-adapted by switching a plurality of filters with different characteristics according to the reduction ratio. In this method, the circuit size becomes large due to the number of filters, so it is not economical and it is difficult to improve the adaptability. Therefore, the number of filters prepared was small and the adaptability was poor.

Second, with conventional circuit configurations, the circuit scale is large, making it difficult to configure high-order filters for improving image quality. In particular, the interpolation filter performs first-order interpolation processing, which causes deterioration in image quality.

Therefore, the purpose of the tree invention is to make the band characteristics of the filter variable without providing multiple filters with different characteristics, so that the characteristics can be easily adapted in real time.
An object of the present invention is to provide a band-limiting interpolation filter that can be easily constructed as a high-order filter in order to improve image quality.

[Means for Solving the Problems] In order to achieve such an object, the present invention includes an input means for manually inputting coordinate transformation data for geometrically transforming an image, and an input means for storing image data. By the integer part of the coordinate transformation data input from the input means, 21'+1 image data P(i') = (i' εi, -1' ,・
...to, ...10+1'), and the 5inc function is equally divided into Tx' pieces up to the first τ point, and the data string sampled at l/2Tx' is divided into 2
A means for storing each in 1'+1 tables,
an enlargement or reduction ratio for geometrically transforming the image;
5inc function first: Refer to 21'+1 tables weighted by the 5inc function using the number of divisions Tx' up to the point and the decimal part of the coordinate transformation data, and calculate 21'
A calculation means for calculating +1 coefficient and a calculation means,
means for extracting, as a filter output, the sum of 21'+1 values calculated from H'+1 data, which are each calculated coefficient, and 21'+1 image data extracted from the extracting means; It is characterized by:

[Function] According to the invention, the impulse response can be freely set by making the tap coefficients of the filter variable. In other words, by adopting a method of referring to a table that stores impulse responses in advance, calculation time for determining tap coefficients can be reduced, and the band characteristics of the filter can be changed simply by changing the table reference interval. It is possible to easily configure an adaptive filter.

Furthermore, by increasing the size of this table, it is possible to easily configure a high-order filter.

Furthermore, by making the impulse response of the band-limiting filter and the impulse response of the interpolation filter the same function, each process can be executed with one filter. [Example] As described above, In FIG. 3, a band-limiting circuit 4 and an interpolation circuit 6 are provided before and after the coordinate transformation processing, respectively, to perform filter processing independently.

In the present invention, these filter processes are performed simultaneously, and the filter can be considered to have comprehensive characteristics.

This method uses a digital-to-digital change t! when the scaling factor is fixed. ! ! (D/D conversion).

However, when changing the expansion/reduction ratio, due to reasons such as the fact that the 11F area 11 limit characteristic is adaptive and the interval of interpolation data changes, conventional methods have been used to obtain the comprehensive characteristic quickly and economically. There was no.

Therefore, in the present invention, by assuming each filter to be an ideal L PF, the calculation for determining the overall characteristic is omitted.

That is, the overall characteristic of a filter is usually obtained by convolving the impulse responses of each filter. However, since the overall characteristics of the ideal LPF are characteristics on the narrowband side, there is no need for calculation, and the impulse response of the overall characteristics is also the impulse response of the ideal LPF.
i in CM p1nx (can be expressed as F(Xl= ).This 5incfi
The numbers are stored as a table, and this is referred to as the filter coefficients.

Bandwidth limiting processing can be made adaptive by changing the interval when referring to this 5inc function. Also,
In the interpolation process, the coefficients on the data to be convolved are determined by moving the decimal part of the coordinates obtained by the coordinate transformation operation past the reference 5 inc function.

Hereinafter, the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of a horizontal filter in one embodiment of the present invention.

The bandlimiting and interpolation filters are spatial filters, which can be composed of a combination of horizontal and vertical filters. Although a one-dimensional horizontal filter will be explained here, it is easy to extend this to two-dimensional filters.

In the figure, IA and IB are input terminals for new coordinate data after coordinate transformation, IA is the decimal part δ, IB
is its integer part i. is input. tV is an input terminal for image data. 2 is an output terminal, from which filtered output image data q (mo) is output.

5 is an image memory, which stores image data from input terminal IV. 12 is a delay circuit (FIFO); 13c is an input terminal with a reduction rate α; 13. is an input terminal with a magnification of 1, Sl
MM is a multiplier which calculates a coefficient that determines an initial value for interpolation of image data.

M↑ is a multiplier that calculates the tap interval according to the scaling ratio. Ll・~Ml・and M−,・′～M,・′ are multipliers, AO,8-8,～8. is an adder, T o ,
T-+ ・~Thi is table, Tx' is 5inch
This is the number of intervals into which the function up to the first τ point is divided.

Next, the operation of the filter of this embodiment will be explained with reference to FIG.

The basic configuration of the filter of this embodiment is a transversal filter type, and the number of taps is H'+1. The integer part of the new coordinate data found by coordinate transformation is 1
0. Let δ be the decimal part. Integer part i. is used as the address of the image memory 5. The image memory 5 stores 21'+1 image data P(i'
) = (i'tEio-1',..., 10...+
, i, +1'). The image data P(i') output from the image memory 5 is delayed by the FIFO 12 by the time necessary to obtain the filter coefficients.

On the other hand, in the tables T-, ~T, ·, the 5inc function up to the first τ point is divided into Tx' pieces, and 1/'27X
' It stores discrete data sampled by '. If the reduction rate is α, the interval to refer to it is Tx when expanding.
', αTx' when contracting, so 1 when enlarging.
, when reducing, α is switched by switch Sl, and multiplier MT
The tap interval is obtained by multiplying by Tx'.

With this value of filter tap spacing, α for each tap is
Tx' is multiplied by 1 (-1' to 1'
) to obtain αTx″i. At the same time, multiplier MM multiplies αTx′ by the fractional part δ of the coordinate transformation data obtained by coordinate transformation and input from input terminal IA to obtain αTx′ δ. Find αTx'(i+δ) by adding them with the adders A-B~A, respectively, and use this to refer to the table T. and T-, , -T, .Refer to the table. By multiplying the tap coefficients of the filter obtained by using the previous image data P(i') using multipliers M0° and M-, , l~-1', respectively, and calculating the sum using adder Ao. Output image q
(mo) is obtained and output to output terminal 2.

[Effects of the Invention] As is clear from the above, according to the present invention, it is possible to configure a band-limiting filter with excellent characteristics that has band-limiting characteristics adapted to the reduction ratio, and eliminates aliasing noise. be able to.

Further, in the interpolation process, interpolation noise can also be removed at the same time because a high-order filter with excellent characteristics is used instead of the conventional linear interpolation.

Furthermore, since band limiting processing and interpolation processing can be implemented with one filter, the circuit scale can be reduced.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment according to the present invention, FIG. 2 is a block diagram showing an example of a conventional video signal scaling device, and FIG. 3 is a conventional video signal geometric conversion device. FIG. 2 is a block diagram showing another example. 1...Input terminal, IA...Input terminal for coordinate data decimal part δ, IB...
Input terminal of coordinate data integer part 10, 2... Output terminal, 3... A/D converter, 4... Spatial filter for band limitation, 5... Image memory, 6... Spatial interpolation filter , 6H...Vertical interpolation filter, 6V...Horizontal interpolation filter, 7...D/A converter, 8...Address generator, 9.10.11...Switch for switching image enlargement/reduction , 12...Delay circuit (FIFO), 13c...Reduction ratio input terminal, 13...Enlargement input terminal, Sl...Switch, MM, MT, M-t~M1・1M1・′ ~Ml
・'... Multiplier, Ao, A-+...A1... Adder, To, T
-+""'T+" Table, Tx'...Number of tap interval coefficients of the filter (pieces).

Claims (1)

[Scope of Claims] Input means for inputting coordinate transformation data for geometrically transforming an image, and storing the data of the image and specifying it by the integer part of the coordinate transformation data input from the input means. 2I′ arranged around the image data i_0 with the number
+ 1 image data P(i′)={i′∈i_0−
I',...i_0,...i_0+I'}, and a data string obtained by equally dividing the Sinc function up to its first zero into Tx' pieces and sampling it at 1/2 Tx', 2I'+1
an enlargement or reduction ratio for geometrically transforming the image, the number of divisions Tx' up to the first zero of the Sinc function, and a decimal part of the coordinate transformation data. Accordingly, the Sinc
a calculation means for calculating 2I'+1 coefficients by referring to each of the 2I'+1 tables weighted by a function; and 2I'+ each coefficient calculated by the calculation means.
1 piece of data and 2I' extracted from the extraction means
1. A band-limiting interpolation filter comprising: means for extracting the sum of 2I'+1 values calculated from +1 image data as a filter output.