JPS5972568A - Picture converter - Google Patents

Abstract

PURPOSE:To increase the density of an input picture element and to prevent the intermittence of output picture elements, by adding a virtual input picture element to an original input picture element. CONSTITUTION:Both original and virtual input picture elements delivered from a virtual picture element adding circuit 3 are supplied to an exclusive hardware 4. The write/read is controlled by a processor 5 for the picture data given from a picture memory 1. Then the processor 5 obtains the writing address of a picture memory 2 from the reading address of the memory 1 and a conversion equation corresponding to a picture conversion pattern. In other words, the processor 5 knows through calculations the area to be converted with an input picture and gives an access instruction of a desired picture element to the memory 1. In this case, the conversion equation uses an equation which performs approximation with a linear conversion, an equation which divides an input picture into many rectangular blocks and at the same time approximates an output picture with many polygons, etc.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image conversion device that performs geometric conversion of an image by digital signal processing.

The present invention is used as a special effects device for broadcasting and an animation creation device.

"Background Art and its Problems" One method of processing a digital video signal to form an image that is a geometrically transformed original image is known as a write address control method. This write address control system will be explained with reference to the IT diagram and FIG. 2.

In Fig. 1, IMl indicates the input image, IM
The mushroom indicates the output image, the pixels of the input image are indicated by white dots, and the pixels of the output image ■M goose are indicated by X marks. These input image IMI and output image IM2 have the same shape, and their respective pixels are arranged at the same position. The position of each pixel in the input image IM is (x, yl)
l! It is represented by I mark, and in the output image IM2,
It is expressed by (X21y2) coordinates.

Then, the input image is scanned, and the pixels of the scanned input image IMI are converted into the output image IM by coordinate transformation.
! The position on the coordinates of is determined. This coordinate transformation is determined by the intended geometric transformation.

The converted pixel of the input pixel IMI is assigned to the pixel closest to the position on the output image IJJz determined by this conversion. In digital signal processing, image memories corresponding to the input image IMI and the output image IMm are provided, the position of one pixel is made to correspond to the address of the image memory, and one image is made into, for example, 8-bit data. Further, in order to perform three-dimensional image transformation, a coordinate axis in the depth direction is also required.

There is also a read address control method as another method of converting one image. This read address control method is based on the input image I
From the coordinates of MI (x+, yl) to the coordinates of output image IM depth (
x2 + y*), extract the pixel of the input image IM that is closest to the position on the input picture image IMI of each pixel of the output image IM2 obtained by this inverse transformation, This can be drawn in each pixel of the output image IM2.

In this way, the write address control method uses the input image IMI
The write address of the output image memory is controlled based on the coordinates of . In addition, the successive address control method is
This is to control the readout address of the input image mesori with reference to the village of the output image IM2. In the read address control method, there is a problem in that the calculation for obtaining the inverse conversion is extremely sloppy (and the hardware required for it is also very expensive). In contrast, the write address control force formula has the advantage of not requiring inverse conversion. However, the write address control method has the drawback that there are pixels in the output image IM2 to which the pixels of the input image IMI are not assigned.

The input image IM1 is set at 45° using the write address control method.
When performing a rotational geometric transformation, the four pixels al v al ta3 of the input image IMI shown in #J2 Figure A
1 a4 is an output image of 1M! In this case, the position is shown in FIG. 2BK. And pixel a after conversion! 4 pixels of the appearance image closest to each of ~a4 bl +1) 1wb3
r b4 is allocated. Therefore, the output image IM
For 20 pixels 1)+sK, there is no image data to be assigned 1. As described above, the write address control method has the disadvantage that output pixels are intermittently produced.

``Object of the Invention'' The present invention is intended to solve problems in image conversion devices using a write address control method. The object of the present invention is to realize an image conversion device that prevents output pixel intermittence. Furthermore, the present invention provides II! processing such as arithmetic operations that can be performed using a pipeline control method. j
The purpose of this invention is to provide an image conversion device.

"Summary of the Invention" This invention forms virtual input pixel signals based on the signals of each pixel of an input image signal according to the enlargement ratio of each part within one screen, and forms each of the input image signals according to a conversion pattern. A pixel signal and a virtual input pixel signal are written to the output side image memory, and each pixel signal and virtual input pixel signal are sequentially read out from the output side image memory to form a converted output image. It is. Furthermore, Jin's invention forms virtual input pixels uniformly over the entire surface of an input image.

"Example" FIG. 3 shows the configuration of one embodiment of the present invention ρl.

An image memory 1 is supplied with input image data and into which one frame of image data is written, and an image memory 2 is into which output image data is written.

Image data sequentially read from the image memory 1 is supplied to a virtual pixel addition circuit 3, and virtual pixels are added thereto.

In this embodiment, the fact that the local magnification of the screen between the input image and the output image is held to a constant value over the entire screen is considered to be a double issue, and the virtual input pixels are uniformly distributed over the entire screen. , in FIG. 4, white dots indicate original input pixels. Fig. 4 U-fI4 type, with 4 pixels vertically and 5 pixels horizontally. And, as shown by the black dot in Fig. 4, 5 K, 'i! The virtual input pixels are added uniformly such that the distance between dims is 1/2 of that for the original input pixels. Therefore, the total number of pixels including original input pixels and virtual human input pixels is (8X10=8
0), which results in an increase of 9.4 times. Such virtual input 1Li11 elements can be formed by linear interpolation using the virtual pixel adding circuit 3.

It is also possible to form the image input pixels at a density corresponding to the local magnification rate. The magnification rate is calculated for each area where several pixels of the input image are read, and the number of virtual input pixels to be added is increased in areas where the magnification rate is large.

This increases the density of input pixels.

The original input pixels and virtual input pixels emerging from the virtual pixel addition circuit 3 are supplied to dedicated hardware 4 .

The writing and continuous output of image data from the image memory 1 is performed by the processor 5, and this image memory 1
The processor 5 determines the write address of the image memory 2 from the read address, the image conversion pattern, and the corresponding conversion formula. The data processor 5 calculates which part of the input image is to be converted to which part and creates a one-image memo IJ.
Give necessary pixel access instructions to IK. As this conversion formula, one that approximates by linear transformation, one that divides the input image into many rectangular blocks and approximates the output image using many polygons, etc. can be used. An input/output device 6 is provided in association with the processor 5, and the type of one image conversion pattern is designated.

The input pixel is folded into the write address of the image memory 2 determined by the dedicated hardware 4. When writing is performed to the same write address twice or more, the latest input pixel is written. This successive output from the image memory 2 is performed in a fixed order, and output image data is taken out from the one image memory 2.

For color images, 1 image memory 1. Virtual l! 11i element addition circuit 3. Dedicated hardware 42 image memory 2 stores each component (Y, U, etc.) of the color video signal.

V).

In addition, since the virtual input pixels are generated uniformly over the entire surface, and the input pixels consisting of the original input pixels and the virtual input pixels are uniformly present, the data is processed using the dedicated node 4 in a pipeline method. can be done. Pipeline processing is a method in which processing data flows through each ring in synchronization with a certain clock. In order to perform this pipeline method, it is necessary that each processing portion of each processing data be processed in the same amount of time. Furthermore, it becomes possible to simplify the hardware and perform high-speed data processing using the pipeline system. As mentioned above, since the number of input pixels is increased four times by adding virtual input pixels, the dedicated hardware 4 needs to process data four times as fast. Therefore, using the pipeline method is effective for meeting such needs.

An example of image conversion is shown in FIG. In this image conversion, an input image IMI consisting of four letters of the alphabet is wrapped around a cylindrical surface without expansion or contraction as shown in FIG. It is something. In this conversion, any two points in the input image are converted.

Considering these two points and the corresponding two points in the output image.

The distance between two points in the output image is never large compared to the distance between two points in the input image.

Further, the input image IMi shown in FIG. 6A is stored as snB.

Even when converting to output images IM2 as shown in FIGS. 6C1 and 6, respectively, no part of the output image is enlarged compared to the input image. As shown in FIGS. 5 and 6, in image conversion actually used, the local magnification ratio is often kept at a constant value of 1 or less.

In this way, when the enlargement ratio of one image is kept at a constant value of 1 or less, K as in this embodiment.

By quadrupling the density of the input image, it is possible to definitely prevent pixel gaps from occurring. This point will be explained with reference to FIG.

FIG. 7A shows input pixels and virtual input pixels of the input image IMI. Assuming that the distance between adjacent input pixels in the horizontal and vertical directions is 0.5.

There must be an original input pixel or a virtual input pixel within a distance of 1- to any point in the input image.

Therefore, as shown in FIGS. 7A and 7B.

If we consider a point 9 in the input image IM1 that corresponds to a pixel 8 in the transformation area 1 of the output image IM, there is a pixel at this point. The distance between this point 9 and the original input pixel or virtual input pixel is h
Therefore, as shown in FIG. 7B, from 1 pixel 8'(<0.
6) If there is always a point in the grain where the original input pixel or the virtual input pixel has been transformed, then it becomes K. Therefore, it is possible to prevent intermittent output pixels such that image data is not assigned to one pixel 8.

FIG. 8 shows another embodiment of the invention. Similar to the embodiment of the present invention shown in FIG. 3 described above, one image memory 1. Virtual il! ii cable addition circuit 3. Dedicated hardware 49 is provided with image memory 2, and 1 access command to image memory 1, il! The processor 5 performs calculations of the write address of the j7 image memory 2 and the like.

Then, the image enlarging device 1 is connected to the image memory 2 on the output side.
0 is set.

This image enlarging device 10 is a device for enlarging an output image written in one image memory 2 at a fixed enlargement ratio of one or more. This image enlarging device 10 as well.

An image conversion device comprising: an image memory for input and output;
Contains clothing in a dedicated bar. This +1! ii The image enlarging device 10 performs simple enlarging of one image,
Its structure is simple and may be of either the write address control method or the successive address control method.

The image conversion performed by the processor 5 and the dedicated hardware 4 is to form a similar output image by reducing the target output image at a fixed ratio. When the output image taken out from the image memory 2 is passed through the image enlarging device 10, a desired output image can be obtained.

In another embodiment of the present invention, even if the local enlargement rate of one image is greater than a certain value, the conversion is performed into a reduced version of the target output image.

Next, since the desired output image is formed by enlarging one image, a method of uniformly generating virtual input pixels can be applied.

"Effects of the Invention" According to this invention, since it is a write address control method,
There is no need to obtain inverse transformation, and software can be simplified. Further, in this invention, the density of input pixels is increased by adding virtual input pixels to original input pixels, so that it is possible to prevent output pixels from being interrupted. Furthermore, by adding a conversion pattern of 9 image conversions or an image enlargement process, if the local enlargement rate of 1 image conversion is suppressed to a constant value, the virtual pixels added by this invention can be made uniform. Therefore, it becomes possible to use the pipeline collateral method as the data processing method.

[Brief explanation of the drawing]

1 and 2 are schematic diagrams used to explain an image conversion device using a conventional write address control method, FIG. 3 is a block diagram showing the configuration of an embodiment of the present invention, and FIG. 4 is a diagram of the present invention. FIGS. 5 and 6 are route maps used to explain virtual input pixels in one embodiment, and FIGS. 6 and 6 are route maps used to explain some examples of image conversion to which this invention can be applied. FIG. The route map used to explain one embodiment, Figure @8, is a block diagram showing the configuration of another embodiment of the present invention. IMl・・・・・・・・・Input both images, 1M2・
・・・・・・・・・・・・Output image, 1゜2・・・・・・
・・・・・・Image memory, 3・・・・・・・・・・・・
Virtual pixel addition circuit. 4...... Dedicated hardware, 5...
・・・・・・・・・Processor. Agent Tadashi Sugiura Tomo B

Claims (2)

[Claims] (1) The input image signal stored in the image memo on the input side is converted into an image memo on the output side according to a conversion pattern of θr constant!
In an image conversion device that incorporates JK4F. A signal of a virtual input pixel is formed based on the signal of each pixel of the input image signal according to the enlargement rate of each part within one screen, and a signal of each pixel of the input image signal and the virtual input are formed according to the conversion pattern. Image conversion in which a pixel signal is written to the image memo IJK on the output side, the signals of each pixel and the signal of the virtual input pixel are sequentially read out from the image memory on the output side, and a converted output image is formed. Device. (2) In an image conversion device that writes an input image signal stored in an image memory on the input side to an image memory on the output side according to a predetermined conversion pattern. A virtual input pixel signal is uniformly formed based on the signal of each pixel of the input image signal, and the signal of each pixel of the input image signal and the signal of the virtual input pixel are transferred to the output side image memory according to the conversion pattern. An image conversion device configured to read out signals of each pixel and signals of the virtual input pixel sequentially from the image memory on the output side to form a converted output image.