TWI281143B - Method for designing video and image scaler based on 2-d finite impulse response filter - Google Patents

Abstract

The present invention is to provide a method for designing a video and image sealer based on 2-D finite impulse response (FIR) filter. First, a 2-D video/image source is sampled with sampling rate higher than Nyquist rate to obtain 2-D samples; second, zeros are padded in between the 2-D samples to get a zero padding image; third, the zero padding image is passed through a 2-D FIR filter and a scaled-up image is obtained.

Description

1281143 V. INSTRUCTION DESCRIPTION α) Technical Field of the Invention The present invention relates to an improvement of an image scaler, and more particularly to designing an image scaler by a 2-D Finite lmpuise Resp〇nse filter. (sc a 1 er ) Method 先前 [Prior Art] Human beings are in a digital world, and digital image processing should be widely used: in many fields, such as personal computers, digital cameras, digital TVs... f & Devices such as LCDs, PDPs, etc. have also become digital display panels of many different sizes on the surface of human life, #影: Source: ί:: ί: fixed. Therefore, it is necessary to use a scaler (the size of the two original display panels). The # Γ=Γ scaler method or interpolation method such as bilinear method or double vertical method is used in related products, but there are some defects. Here are some traditional methods:

(1) Copy method please refer to the figure _ • Just copy the previous figure 2, where C and D are three times, this method is inserted, and the result is shown in '# size must be enlarged to the origin (such as A or B) and Insert as an interior point is a copy of point A. (2) Bilinear method In this method, C and η machines are π-port Λ -, which account for the replication of the point, but the combination of point a and point β.

Page 6 1281143 V. Description of invention (2) Cheng. Since point C is closer to point a, point c has more components of point A. On the other hand, point D has more components of point b. Take linear synthesis and get:

C = 2/3*A + 1/3*B

D = 1/3*A + 2/3*B The results are shown in Figure 3. (3) The double vertical method, this method is similar to the bilinear method, except that the synthetic equation is a cubic equation. The image is zoomed ‘and then the image, ^. When scaling the image, the above method is applied to the horizontal sample and also to the vertical sample of the result, so the final image is shown in Figure 4. The above-mentioned scaling or interpolation method is widely used in different collar=copying methods. 'The image scaler is out of the edge of the screen. Although the double-linear method is used, the surface is blurred; *Using the double-jj point It’s not clear enough for the human eye. No matter which method is used, the two are operated in the horizontal and vertical directions, as shown in Figure 4. No: All ”images are basically two-dimensional, s this one-dimensional (but natural...the result of the image will be distorted, it’s not visible to the human eye): put

1281143

However, the object of the present invention is to improve and/or eliminate the above disadvantages. SUMMARY OF THE INVENTION A primary object of the present invention is to provide a method for designing an image scaler with a rare wave-wave device. First, G,, 'Z impulse response 遽Γ! ίί lst rate) is sampled to obtain a two-dimensional transcript, and secondly, zero is filled between the two-dimensional samples to obtain a zero-fill image, the third step , the zero-fill image pass:, the response filter (2-D FIRfilter): the dimension of the finite pulse - Yu Wei to a magnified image. The required magnification is L, and the required image processing block size 疋Mx N, then insert the zero point between the two-dimensional samples of N , to get a L) ^ XL) X (NXL) zero fill Input image; take a (MxL) X (NXL) two-dimensional finite impulse response filter; the (M xL) X (Nx L) 4 one-dimensional finite impulse response filter has (Μχ L) χ (NxL) (MxL) x (NxL) zero-fill image is processed by the corresponding coefficient of the (MxL) χ (Νχ two-dimensional finite impulse response filter, and then added to become (Μ XL) X ( Ν XL) the result of the image processing of the center pixel of the block. The image is repeated until the next image to be processed is processed until the position of the next image is processed. [Embodiment] The present invention proposes A method of designing an image scaler using a two-dimensional finite impulse response filter.

Page 8 1281143 V. Invention Description (4) " ---- Please refer to Figure 5, the first step is to fill the zero point between the original samples; the - under the zero point fill image through a two-dimensional limited The impulse response filter then produces an amplified image. If the size of the enlarged image is not the original image = integer multiple, or smaller than the original image, the original image can be enlarged to an integral multiple of the final image, and then part of the sample is taken as the final image. According to the sampling theory of digital signal processing, the digitized signal can be restored to the original analog signal only when the sampling rate is higher than the Nyquist rate. The signal spectrum obtained by filtering out the high frequency band is completely the same as the original spectrum. The lower the sampling frequency is, the more the low pass chopper is required, for example, the higher the Nyquist rate than the Nyquist rate. The sampled signal obtained, Figure 7 is its spectrum. In order to obtain an N-fold shadow f, (N-1) zeros are filled between the original samples, as shown in Fig. 8. The graph is the spectrum after filling in the zero point. If the filter 1 other than pi/N: pi/N] is trusted, the new spectrum will be obtained as shown in Fig. 10. #图六六样本率 I / mouth attendance to sampling 仏纟, as shown in Figure -11, then the spectrum and graph

^ Same. This proves that the zero fill is the same as the improvement of the original signal. The t' impulse response is based on the chopper, which means that it is not necessary to be imaged. Two-dimensional finite impulse response filtering is decomposed. Since the filter must be an image, the two-dimensional finite impulse response is based on a one-dimensional finite pulse. The coefficient of the two-dimensional finite impulse response in the horizontal and vertical directions can be decomposed or non-linear. To avoid blurry shadows

Page 9 1281143 V. INSTRUCTIONS (5) The coefficients of the filter are symmetrical with respect to the center point. The method of finite impulse response filter is the same as that of the one-dimensional finite impulse response chopper. The zero point is inserted between all the original samples to achieve the required size, as shown in Figure 12. Then, the spectrum of the zero-filled image is filtered by a two-dimensional finite impulse response filter, and the spectrum of the obtained image has the same spectrum as the oversampled sampled signal. Resize the enlarged image and take the aliquot of the desired sample as the final image. Implementation of Jj and scaler related products for personal computer LCD monitors, LCD TVs, personal digital assistants (PDAs), etc. For economical benefit, these products are implemented in microchip or integrated circuits. The following explains how to design a scaler' that can be easily implemented with VHDL/Verilog RTL hardware code. Referring to Figure 13, the first step is to store the original image to the memory. Select the desired horizontal and vertical magnification, called L;

The image processing block size of the pulse-limited response wave is called Μ X N ; the zero point is inserted into the Μ X Ν image (the shaded area in Figure 13), and a zero-filled UxL) X (Nx L) image is obtained. Then a (MX L) X (NxL) dimensional finite impulse response filter is used. The coefficients of the two-dimensional finite impulse response filter are shown in the upper right and lower corners of Figure 13: { F —〇一〇, F —1 —0, F - 2 - 〇, ... · F - Mx L 〇 F - 〇 - 1, F - 1 - 1, F - 2 - 1, ..., F - Mx xL",

Page 10 1281143 V. Description of invention (6) L,···· F —M x L — N x L }

F.N x L, F-1J x L, Filtering In order to obtain a zero-filled impulse response, the waver passes through the image. Zero Vm must be processed with two-dimensional finite veins with corresponding coefficients, and added to form L) X (NxL) images (as shown in Figure 13, the result of the center pixel is the center of the T-shirt area. A position to be processed, repeat A does not; move the image block output, so repeat until all the soils get the next pixel to finish 'then get - (Mx L) x (NVu-type pixels are processed to output image storage Memory, and select the enlarged image of J XL); this is compared with the output of the vertical size f. The pixel achieves the desired level. The advantages of the present invention are as follows: The method can be used to make the final number of long filters, wave squares, finite impulses, or not C. The present invention proposes a coefficient-based: Parameter method to implement the scaler and interpolator

The impulse responds to the filter coefficients to control the image quality. ‘Easy > Month is clear or smooth. Two-dimensional finite impulse response filtering can be adjusted for different design requirements. The longer the M W coefficient is, the better the resulting image. The filter is realistic.

Page 11:: One advantage is that the present invention proposes a two-dimensional finite impulse sound image processing method, # abandoning the past one-dimensional image processing. 1281143 V. INSTRUCTIONS (7) The world's images are two-dimensional, and processing images in one-dimensional direction loses image quality. Because of the above advantages, the present invention is indeed an innovative and superior method for designing scalers, interposers, and other image processing applications. While the present invention has been described and illustrated by the present invention, it will be apparent to those skilled in the

Page 12 1281143 Brief description of the diagram [Simple description of the diagram] Figure 1 is a schematic diagram of enlarging the signal to three times as large as the original. Figure 2 is a schematic diagram of the replication method. Figure 3 is a schematic diagram of the bilinear method. Figure 4 is a schematic illustration of the scaler operating linearly in the horizontal and vertical directions, respectively. Figure 5 is a schematic illustration of the process of the invention. Figure / shows a sampled signal at a higher Nyquist rate. • Figure 7 shows the spectrum of the sampled signal in Figure 6. Figure 8 shows (N-丨) zero points filled in between the original sampled signals and an image of N times. Figure IX shows the spectrum after the sample signal is filled in zero. After filtering out the signal in the frequency band other than [_pl/N: pi/N], the result is 4:;: If the sampling signal is sampled at the sampling rate of _ times, the spectrum of the signal and the spectrum of Fig. 10 are obtained. same. Dot plug: = Limit pulse response In the wave method, zero > Η , and final sampling 仏 5 between the tigers to obtain a desired magnification. Figure 12 shows how the invention can be implemented. Explanation of main component symbols] None (the symbols in the figure are not representative components, but are used to guide instructions

Page 13

Claims (1)

1281143 VI. Patent Application 1 · The method of designing an image scaler with a two-dimensional finite impulse response filter includes the following steps: a• Sampling a 2D image source at a higher Nyquist rate And get a two-dimensional sample·, b· fill the zero point between the two-dimensional samples to get a filled image; suppress c·fill the zero-fill image through a two-dimensional finite impulse response filter 2: two R fUter ) 'So get - magnify the image. k and: I = profit range The first item is designed with a two-dimensional finite impulse response filter. The method of the shirt imager is straight Φ. It is w, and the magnification required for the right is L·, And, if the size of the .t square is M x N, insert the zero point into a two-dimensional X of XN to obtain a zero-fill image of (MxL) x (Nxl); (NxL) two-dimensional finite impulse response filter; X's two-dimensional finite impulse response filter has (Mx x UN XL) coefficients; (Nxtr six-zero fill-in image with this (MxL) X ίίΪΐ (MXL Χ (NXL) image processing block center pixel j • Move the image processing block to the next position to be processed, repeat the above sequence until all the zero-fill images are processed.