KR100994770B1 - Picture size conversion apparatus using multi directional filtering - Google Patents

Abstract

Disclosed is an image size conversion apparatus for obtaining a size converted image with more perfect image quality by considering size conversion filtering on components other than the vertical and horizontal directions. The image size converting apparatus extracts and outputs at least one one-dimensional directional signal for a direction other than the vertical direction and the horizontal direction from the input image, and synthesizes each pixel value of the input image into the value of the one-dimensional directional signal. A decomposition converter for generating and outputting a residual signal from a difference between a signal and the signal of the input image, a directional signal filter for filtering the one-dimensional directional signal output from the decomposition converter to correspond to a predetermined size, and the decomposition converter And a synthesizer converter for synthesizing the residual signal output from the filter to correspond to the predetermined magnitude, and synthesizing the directional signal and the residual signal filtered by the directional signal filter and the residual signal filter. In addition, by using multiple one-dimensional directional signals, the final residual signal can be made less than a predetermined reference energy, thereby reducing the filtering effort for the residual signal.

Image size conversion, filtering, one-dimensional directional signal

Description

Picture size conversion apparatus using multi directional filtering}

1 is a block diagram of a conventional image size conversion device.

2 is a view for explaining a conventional image size conversion process.

3 is a block diagram of an image size converting apparatus according to a preferred embodiment of the present invention.

4 is a view for explaining a directional signal according to a preferred embodiment of the present invention.

5 is a diagram for explaining decomposition, synthesis, and generation of a residual signal of a directional signal according to an exemplary embodiment of the present invention.

The present invention relates to an apparatus for converting an image size, and more particularly, to an apparatus for converting an image size in consideration of signal components in other directions as well as signal components in the horizontal and vertical directions of the image.

In recent years, with the spread of digital broadcasting and the Internet, a so-called transcoding technique for changing the resolution of compressed images has been important. This transcoding technology is, for example, a screen division display in which images of various broadcast programs received in digital broadcasting are reduced and synthesized and displayed on a TV receiver as one stream, or an image in any compressed format is compressed in a compression format for a TV conference. It can be used for a format conversion for changing and converting, an image relay device for controlling bit-rate by changing the resolution of a compressed image in accordance with a network bandwidth or a processing capability of a computer. In addition, the transcoding technique can be applied not only to moving pictures but also to still pictures.

FIG. 1 shows an example of a conventional moving picture transmission apparatus which reduces a size of a moving picture and transmits it. The transmitting device 10 includes an A / D converter 12 for converting an input analog moving picture into a digital signal, a horizontal filter 14 for converting the horizontal size of the original image, and a vertical for converting the vertical size of the original image. And a direction filter 16 and an encoder 18. The encoder 18 performs a predetermined encoding process on the size-converted original image, and simultaneously converts the image coded data thus obtained into MPEG2 (Moving Picture Experts Group Phase 2) standard. Transport stream packet data obtained by transport stream packetizing according to the format specified in FIG. At this time, when the input original image is a digital image, the A / D converter 12 is not necessary. In addition, the size of the original image may be converted by the image receiving apparatus. In this case, the size of the original image may be converted by decoding the encoded original image signal.

FIG. 2 illustrates a process in which the vertical component pixels and the horizontal component pixels of the original image are size-converted by the horizontal filter 14 and the vertical filter 16 of FIG. 1. According to FIG. 2, an original image divided into a plurality of horizontal one-dimensional pixel signals is obtained by converting the size of each horizontal pixel signal to obtain an image whose size is converted in the horizontal direction. The image is separated into a plurality of vertical one-dimensional pixel signals, and the magnitude of each vertical pixel signal is converted to obtain an image having a final converted size. At this time, converting the size of the pixel signal means that if the one-dimensional signal is composed of five numbers, for example, the number is changed to three or eight, and the converted signals are similar in shape to the existing signals. Should be done.

In addition, although only a filter is illustrated in the drawing, the detailed process of converting the magnitude of the one-dimensional signal is performed through upsampling, downsmmpling, and filtering, for example, to increase the size of the signal. In some cases (so-called interpolation), upsample the existing signal and then perform interpolation filtering. To reduce the size of the signal (so-called decimation), you can decimate the existing signal before downsampling. At this time, upsampling supplements a signal having a zero value between existing signals, and downsampling discards some of the existing signals.

However, since the conventional image size conversion technology filters only the vertical and horizontal directions when converting the size of the original image, only the signal components appearing in the vertical and horizontal directions among the features of the image are considered. In the converted image, there is a problem in that components in other directions such as diagonal directions of the original image cannot be represented effectively.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object of the present invention is to maintain the image quality of an original image by performing size conversion not only on signal components in the horizontal and vertical directions of the input image but also on signal components in other directions. An image size converting apparatus capable of converting an image size while maintaining it is provided.

In order to achieve the above object, the present invention provides an image size conversion device for converting the size of an input image, by extracting at least one one-dimensional directional signal for a direction other than the vertical direction and the horizontal direction from the input image A decomposition converter for outputting and generating a residual signal from a difference between a signal obtained by synthesizing each pixel value of the input image into a value of the one-dimensional directional signal and the input image, and the outputted from the decomposition converter. A directional signal filter for filtering a one-dimensional directional signal to correspond to a predetermined magnitude, a residual signal filter for filtering the residual signal output from the decomposition converter to correspond to the predetermined magnitude, the directional signal filter and the residual signal A synthesizer that synthesizes and outputs the filtered directional and residual signals An image size converting apparatus is provided.

At this time, the decomposition converter generates a final residual signal by repeating a process of generating a next residual signal a predetermined number of times from a difference between a previous residual signal and a synthesized signal synthesized from the one-dimensional directional signal extracted from the previous residual signal. The residual signal filter filters the final residual signal, and the one-dimensional directional signal may be extracted by overlapping the same directional signal.

The one-dimensional directional signal may be configured by average values of pixel values included in a corresponding direction.

In addition, the residual signal filter may omit the residual signal filtering when the energy of the final residual signal is less than a predetermined value.

In addition, the one-dimensional directional signal may include a one-dimensional directional signal in the vertical direction and / or a one-dimensional directional signal in the horizontal direction.

Also, an area divider for dividing the input image into a plurality of areas having a predetermined size, and a region combiner for combining the synthesized signals for the plurality of areas output from the synthesis converter and outputting an image having the final size change. It is preferable to further include.

Hereinafter, with reference to the accompanying drawings will be described a preferred embodiment of the present invention.

3 shows an image size converting apparatus according to the present embodiment. The image size converting apparatus 30 is composed of an area divider 31, a decomposition converter 32, a directional signal filter 33, a residual signal filter 34, a synthesis converter 35 and an area combiner 36.

The area divider 31 divides the original image input to the device 30 into several areas as necessary. Therefore, the area divider 31 may be omitted when the area division is not necessary.

이다. 입력 영상을 방향성 신호로 분해하는 과정을 F라고 하고, 그 분해된 방향성 신호를 합성하는 역과정을

라고 한다면 입력 영상의 잔여 신호는 다음의 [식 1]과 같다.The decomposition converter 32 sequentially receives the divided regions in the region divider 31 and decomposes the image X of each region into k directional signals Ci and residual signals Cx, y. At this time, the directional signal Ci is a signal consisting of Ni coefficients,

to be. The process of decomposing the input image into directional signals is called F, and the reverse process of synthesizing the decomposed directional signals is referred to.

In this case, the residual signal of the input image is shown in Equation 1 below.

[Equation 1]

For example, the decomposition converter 32 decomposes and outputs the k directional signals from the input image, and also outputs the residual signals, which are 0 degrees, 90 degrees, ± 45 degrees, and ± 22.5 based on the horizontal direction. Assume that the degree, ± 67.5 degrees and so on. At this time, these angles may be used several times in the filtering process, and the total used directional number k is a number including all such overlaps.

이다. 따라서, 도 4의 (a) 0도의 경우에는 Ni=8이 되고,

={p1, p2, ...,p8}이 된다. A method of obtaining a directional signal for one directional will be described with reference to FIG. 4. 4 shows (a) 0 degrees (b) 45 degrees (c) 22.5 degrees of directionality. For example, in FIG. 4 (a), the average value of eight pixels arranged in the horizontal direction is horizontal to the line. It becomes a direction signal. That is, the one-dimensional signal from p1 to p8 becomes the one-dimensional directional signal in the horizontal direction of the 8 * 8 block. Also in the case of (b) and (c) of FIG. 4, one-dimensional directional signals from p1 to p15 or p1 to p11 are similarly obtained. As such, it can be seen that the number of one-dimensional signals in each direction may vary according to the directionality. Each directional signal Ci is a signal consisting of Ni coefficients.

to be. Therefore, in the case of 0 (a) of FIG. 4, Ni = 8,

= {p1, p2, ..., p8}.

Creating a directional signal decomposed in one direction is as described above, and restoring (synthesizing) the one-dimensional signal again by copying each coefficient of the directional signal in the used direction. That is, in the case of (a) 0 of FIG. 4, all eight pixels of the top row have a value of p1, the second row has a value of p2, and the last row has a value of p8.

Referring to Figure 5, the decomposition and synthesis of the directional signal will be described as follows. The composite image of the original image and the one-dimensional signal will generally not be the same. This is because, in a special case, for example, if the pixel values of one line are all the same and their average value is the same, the synthesized image of the original image and the one-dimensional signal will be the same, but generally it will not. Therefore, the difference between the original image and the synthesized image is called a residual signal. When the second directional signal is calculated and outputted with respect to the first residual signal, a difference between the signal synthesized with the second directional signal and the first residual signal occurs. In this case, this second difference is referred to as a second residual signal. The kth final residual signal generated after k steps of this process becomes the residual signal Cx, y in the present invention. This is expressed as a formula above [Equation 1].

가 된다. 이 때, 영상의 경계 부근을 필터링 할 때에는 주변 영역의 영상으로부터 나온 방향성 신호도 이용하여 처리하는 것이 바람직하다. 또한, 필터링 과정은 종래와 마찬가지로 적절한 크기로 변환하기 위한 업 샘플링 또는 다운 샘플링 과정을 거친다.In the directional signal filter 33, each directional signal Ci is filtered for each filter Tifh to change the value and number of signal coefficients. That is, each filtered directional signal

Becomes In this case, when filtering the vicinity of the boundary of the image, it is preferable to process by using the directional signal from the image of the peripheral area. In addition, the filtering process goes through an upsampling or downsampling process to convert to an appropriate size as in the prior art.

The residual signal filter 34 converts the residual signal Cx, y into a residual signal size using a conventional general image filtering method. On the other hand, if the energy of the residual signal (eg, the mean square value of the residual signal pixels) is not greater than a predetermined reference, the residual signal may be discarded and this process may not be performed. For example, if the residual signal that does not significantly affect the human vision can be discarded, it does not matter much, thereby reducing the computational complexity and speeding up the process. On the other hand, as shown in FIG. 5, as the one-dimensional directional signal is repeatedly used several times, the probability of the energy of the final residual signal is smaller.

로 표현할 수 있다.The synthesis converter 35 synthesizes and converts each directional signal passed through the filtering, and when there is a residual signal that is not discarded by the residual signal filter 34, the filtering residual signal C'x, y, which filtered the residual signal, is added. Finally, the image X 'whose size is converted is obtained. In other words,

.

로 표현할 수 있다.In the above equation, it is expressed as if the reconstructed image is made by the interaction of each directional signal.However, if you use the decomposition method described above (that is, obtaining the average value of the directions) to produce the directional signal, it represents The above equation may be expressed as the sum of the synthesized signals of each directional signal. In other words,

.

When the original image is divided into several regions of a small unit in the region divider 31, the region combiner 36 combines the regions which have been separated, filtered, and synthesized, and outputs an image which is finally converted in size.

As described above, according to the present invention, by performing size conversion not only on the signal components in the horizontal and vertical directions of the input image but also on the signal components in other directions, the size of the image can be converted while reproducing the image quality of the original image more completely. It can be effective.

Claims (7)

An image size converting apparatus for converting a magnitude of an input image signal, From the input image signal, a plurality of one-dimensional directional signals in a plurality of directions including a vertical direction and a horizontal direction are extracted and output, and the residual from the difference between the signal obtained by combining the plurality of one-dimensional directional signals and the input image signal. A decomposition converter for generating and outputting a signal; A directional signal filter for filtering each of the plurality of one-dimensional directional signals output from the decomposition converter to correspond to a preset magnitude; A residual signal filter for filtering the residual signal output from the decomposition converter to correspond to the preset magnitude; And a synthesis converter configured to synthesize the directional signal and the residual signal filtered by the directional signal filter and the residual signal filter. The final residual signal of claim 1, wherein the decomposition converter repeats a process of generating a next residual signal from a difference between a previous residual signal and a synthesized signal synthesized from the one-dimensional directional signal extracted from the previous residual signal. And outputting the generated residual signal, wherein the residual signal filter filters the final residual signal. The apparatus of claim 2, wherein the one-dimensional directional signal can be extracted by overlapping the same directional signal. The image size converting apparatus of claim 1, wherein the one-dimensional directional signal is configured by average values of pixel values included in a corresponding direction. The apparatus of claim 1, wherein the residual signal filter omits the residual signal filtering when the energy of the residual signal is less than a preset value. The image size converting apparatus according to claim 1, wherein the one-dimensional directional signal comprises a one-dimensional directional signal in a vertical direction and a one-dimensional directional signal in a horizontal direction. The image of claim 1, further comprising: a region divider for dividing the input image signal into a plurality of regions having a preset size, and a combination of the composite signals for the plurality of regions output from the synthesis converter to finally convert the image. And an area combiner for outputting an image.

Images