KR0164770B1 - Data selection circuit for image compression in three dimension subband coding system - Google Patents

Abstract

A frequency band signal selection method for image compression in a three-dimensional subband encoded image processing, comprising: converting Y, I, and Q signals to RGB image signals obtained from an image acquisition medium, and converting the converted Y, I, and Q signals into a space The signal is divided into signals having a total of n frequency components in a predetermined band on the (vertical and horizontal) time axis.

Among the n subbands divided above, at least the low frequency band which plays an important role in a given image is always transmitted, and the subbands are divided into subblocks in order to transmit components of the band except the low frequency band to a channel having a bandwidth of 6 MHz. Then select and send in order of having high energy.

Description

Data Selection Circuit for Image Compression in 3D Subband Encoding Image System

1 is a system diagram for image compression in a three-dimensional subband encoded image system according to the present invention.

2 is a diagram illustrating the division of each of the Y, I, and Q components of a transform / subband 8x8x3 band according to the present invention.

3 is an MSB search adaptation selection flowchart according to the present invention.

4 is a diagram illustrating a 32-sample sub-block division of each band 160 x 90 according to the present invention.

5 is a diagram showing a difference between an MSB verification method and an optimal adaptation selection method according to the present invention.

6 is an exemplary view showing an MSB verification procedure according to the present invention.

The present invention relates to an efficient frequency band signal selection circuit for compressing image information in an image processing system. Particularly, in the MIT-CC method, an image signal is divided into frequency components by a subband analysis filter and then the magnitude of each frequency energy component. The present invention relates to a data selection circuit for image compression in a 3D subband image encoding system that compresses data by selecting a large energy value by irradiating.

HDTV (High Definition Television) for obtaining large screen and high-definition images has been widely developed in each country since it was developed by NHK in Japan in 1964 due to its wide range of applications in terms of technology as well as its huge impact on other fields. Has been studied. According to the International Radio Consultive Commitee (IRCC), the HDTV is known to have a horizontal, vertical resolution and an aspect ratio of 5: 3 or 16: 9 more than twice that of current color TV systems. In addition, it solves cross color, cross luminance, and ghost phenomena, which are factors of image quality degradation, and provides sound quality comparable to CD (Compact Disk). In order to satisfy the above conditions, the amount of video information to be processed in the HDTV is increased by about five times compared to the current method. Therefore, the research direction of the ATDT method in the US requires an efficient video information compression process to maintain high quality image while transmitting HDTV signal at 6MHz, which is the bandwidth of the current TV. Therefore, in addition to the research on the image compression method proposed in the conventional HDTV, it is very necessary to develop its own technology. In addition, the development of the image compression technology as described above has a great ripple effect in the related digital communication, digital signal processing field. Until now, the HDTV method proposed by the FCC as a whole for terrestrial broadcasting is largely incompatible with the current RC receiver and the RC (Receiver Compatible: RC) method, which is not compatible with the current receiver, but uses the 6MHz bandwidth to transmit channel compatibility. It can be classified by the CC (Channel Compatible: CC) method. Currently, most of the methods used as encoding processes for compressing image information are to analyze image information through a transform or subband and select only important components according to the data amount of a given system. Thereafter, the selected information is decoded. However, the method of selecting an important component in the transform domain is variable according to the compression of the image information so that the energy is selected in order from the signal having a large energy until the data amount of a given system is reached. The method of sequentially selecting the energy from the large signal is the best for restoring the video information, but in order to do this, the size of the entire reproduction signal must be sorted in order. In addition, in the case of HDTV, there is a problem that the alignment of the entire target to be compressed becomes very large.

Accordingly, an object of the present invention is to provide an image compression circuit using subband coding in MIT-CC, which is a channel compatibility scheme proposed by MIT.

Another object of the present invention is to provide a circuit capable of efficiently compressing by separating a frequency component by a subband analysis filter and selecting a large energy value by investigating the size in units of frames of each frequency component.

In order to accomplish the above object, the present invention provides a frequency band signal selection circuit for image compression in three-dimensional subband encoded image processing, comprising: matrix conversion for converting an RGB image signal obtained from an image acquisition medium into Y, I, and Q signals. And a transform / subband analysis process of dividing the Y, I, and Q signals converted in the matrix transformation process into signals having a total of n frequency components in predetermined bands on a spatial (vertical and horizontal) time axis; Of the n subbands divided in the form / subband analysis process, at least the low frequency band, which plays an important role in a given image, is always transmitted, and again to transmit the components of the band except the low frequency band to a channel having a bandwidth of 6 MHz. After dividing into sub-blocks, the adaptive selection process consists of adaptive selection in order of high energy.

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

1 is a system diagram for image compression in a three-dimensional sub-band encoded image system according to the present invention. The matrix converter 101 converts Y, I, and Q signals from an RGB image signal obtained from an image acquisition medium, and the matrix. A three-dimensional subband analyzer 103 for dividing the Y, I, and Q signals converted by the converter 101 into signals having a total of n frequency components in a predetermined band in three dimensions of a spatial (vertical, horizontal) time axis, and Among the n subbands divided by the 3D subband analyzer 103, at least the low frequency band which plays an important role of a given image is always transmitted, and the components of the band excluding the low frequency band are transmitted to the channel having a bandwidth of 6MHZ. The adaptive selector 105 which divides the data into sub-blocks and selects them in order of having high energy, and the adaptive modulator 1 which adaptively modulates the output of the adaptive selector 105. 07) and. The output of the scrambler 109 and the output of the adaptive modulator 107 and the output of the scrambler 109 and the additional data, the output of the digital audio stage scrambler 109 to scramble the output of the adaptive modulator 107 It consists of a multiplexing modulator 111 for multiplexing.

2 is an exemplary diagram of 8x8x3 band division in each of the Y, I, and Q components according to the present invention.

3 is an MSB search adaptation selection flowchart in the adaptation selector 105 of FIG. 1 according to the present invention. The total bit and selection subblock indexes representing the total number of subblocks to be processed and the value for each block and the MSB selection to be processed are shown in FIG. An initializing process for specifying an initial value for the MSB, an MSB checking process for checking the state of the MSB from the first subblock selected, and an average band for reading the previous selected subblock index value when the MSB value is present in the MSB checking process The number of blocks to be searched for MSB is searched when the selected subblock index searching process checks whether the number is reached, and when the MSB value is not found during the MSB checking process and the corresponding average number of bands is not reached during the selection subblock index search process. When the number of blocks to be searched is reached in the process of checking whether to increase the total block search and checking the whether to search the entire block, A search is made to the MSB bit specifying process of specifying a bit for checking the MSB.

4 is a diagram illustrating a 32-sample sub-block division of each band (160 × 90) according to the present invention, and FIG. 5 shows a difference between the MSB verification method and the optimal adaptive selection method according to the present invention. Exemplary diagram showing the MSB verification procedure according to the present invention.

Accordingly, the present invention provides a transmission matrix converter 101 for converting Y, I, Q signals from an RGB image signal obtained from an image acquisition medium, and the Y, I, Q signals converted from the matrix converter 101 in space (vertical, Horizontal 3D subband analyzer 103 for dividing a signal having a total of n frequency components into a predetermined band in three dimensions of the time axis, and a given image of the n subbands divided by the 3D subband analyzer 103. An adaptive selector that transmits at least the low frequency band which plays an important role at all times and divides the subbands except the low frequency band into sub-blocks in order to transmit them to a channel having a bandwidth of 6 MHz. 105, an adaptive modulator 107 for adaptively modulating the output of the adaptive selector 105, and a scrambler 109 that scrambles the output of the adaptive modulator 107. From the system consisting of the multiplexing modulator 111 which multiplexes the output of the adaptive selector 105, the output of the adaptive modulator 107, the output of the scrambling unit 109, the additional data, and the output of the digital audio stage, In the method for image compression through the adaptive modulator 107, the scrambling 109, and the multiplexing modulator 111, a given image of the n subbands analyzed by the three-dimensional subband analyzer 103 is obtained. The first step of selecting the signal of the important frequency and low frequency band in the adaptive selector 105 at all times, modulated in the adaptive modulator 107, scrambled in the scrambler 109 and transmitted to the multiplexing modulator 111, and In the first process, the component of the low frequency band transmitted directly from the adaptation selector 105 is divided into sub-blocks to be transmitted to a channel having a bandwidth of 6 MHz. After the selection in the order of having a high energy, the modulator 107 is modulated by the adaptive modulator 107 and scrambled by the scrambling 109 to be transmitted to the multiplexing modulator 111.

Therefore, an embodiment of the present invention will be described in detail with reference to FIGS. 1 to 6, where a high resolution RGB image signal (eg, 720 line / frame, 1280 pel / line) obtained from a TV camera or other medium in FIG. 60 frames / sec) are converted into Y, I, and Q signals through conversion in the transmission matrix converter 100. The Y, I, and Q signals are transformed by a 3D subband analyzer 103 and subjected to a transform / subband analysis. The signals are composed of 8 bands on the spatial frequency axis and 3 bands on the time frequency axis. It is divided into a signal having a total of 8x8x3 = 192 frequency components. The twelve low frequency signals (four components each of Y, I, and Q), which play the most important role in constructing a given image among these 192 frequency components in the adaptive selector 105, for the phased subband, are always transmitted. I. Q component signals are not transmitted. The Y component signals of the remaining 188 bands are divided into subblocks and then adaptively selected and transmitted in order of high energy in order to be transmitted to a channel having a bandwidth of 6MHZ. The signal selected as described above is modulated in the adaptive modulator 107 to reduce the influence of channel noise, and is transmitted to multiplex in the multiplexing modulator 111 with digital data scrambled in the scrambler 109 to represent information. The receiving end performs the reverse process at the transmitting end to reproduce the original image.

In detail, the 3D subband analyzer 103 describes a transform and subband analysis process. In the present invention, the MIT (Massachusets Institute Technology) transmits an HDTV image information signal through a channel having a 6MHZ bandwidth and the current TV. A vertical (V), horizontal (H), and time axis filter of the Y, I, and Q signals converted by the transmission matrix converter 101 by the MIT CC (Channel Compatible) scheme are proposed in the 3D subband analyzer 103. In the spatial frequency V and H regions, the signal is divided into three bands in the 8 x 8-hour frequency region as shown in FIG. The division into three bands in the time axis region is as follows.

That is, the 3-point DCT from 1 is divided into 8x8 bands in the spatial frequency H and V regions by a quadrature mirror filter (QMF) bank. The 3-point DCT was revealed in 1990 by KR Rao P Yip in Discrete Consine Transform its Application, and the QMF bank theory was published in 1976 by A. Croisier, D. Esteban, and C. Galand: Perfect channel splitting by use of interpoplation / decimation / tree decomposition techniques, by International Conf. on Information Science and Systems. Patras. Presented in Greece and published in May 1977 by D.Esteban and C.Galand in the Application of quadrature mirror filters to split band voice coding schemes. of the IEEE International Conf. on ASSP. Hartford. Connecticut. pp. Published in 191-195. Therefore, each component of Y, I, and Q is divided into 8 bands in the spatial frequency domain and 3 bands in the temporal frequency domain by a transform / subband analysis filter. The horizontal and vertical spatial domain filter has 16 tabs and the time domain filter uses a detachable filter with three tabs. For the transform / subband analysis, three frames of Y, I, and Q signals should be stored, and each of Y, I, and Q signals has 8 × 8 × 3 = 192 frequency components as shown in FIG. Each band signal has 1280/8 x 720/8 = 160 x 90 signal values / band signal values. The channel having a bandwidth of 6MHZ per second using RF-Schreiber et, al., Noise-MARGIN method published in SNPTE Journal as A Compatible High-definition Television System Using the Noise-Margin Method of Hiding Znhancemant Information in December 1988. 10M analog values and 10M digital bits can be transmitted. In the present invention, since three frames are analyzed at the same time, 0.5M analog values and 0.5M digital bits can be transmitted per 1/20 second. All components corresponding to four frequency bands that play the most important role in constructing a given image in each of the divided Y, I and Q frequency domains are selected. In the case of the Y component, the lowest 4 frequency bands are selected from the I and Q components by investigating the energy of the I component of each frequency band. No further selection (160 x 90) x 4 bands x 3 (YIQ) = 1,721,800 analog values). The Y component (60 + 64 + 64) × (160 × 90) = 2,707,200 analog values of the remaining 188 frequency bands is divided into subblocks so that the sum of the absolute values in each subblock is selected in order of large order (327,200 analog values). , 12%), which indicates two-position information, which is 84,600 digital bits. That is, about 12% of data are selected. The image in space-time is divided into components of 192 frequency bands, and since four frequency band components have already been selected, they correspond to the same position in space as shown in FIG. 4, but exist in different sub-blocks in 188 frequency bands. Therefore, after arranging sums of absolute values of subblocks located at the same position in space, subblocks of an average 2217 (22 to 23) band corresponding to 12% of them can be selected. However, in the case of the selection as described above, the complexity of each image corresponding to each subblock in the space is different, and when a constant 12% is selected in the frequency band, the energy of the subblock discarded in the subblock at one position is selected in the subblock at another position. There may be a case larger than the energy of the subblock. Therefore, after all the blocks in the 188 bands are arranged in the order of the sum of the absolute values of the subblocks, selecting a subblock corresponding to 12% from the subblock with the highest absolute value results in a much better result in the image restoration. . However, the number of bits for enumerating the absolute sums of the total subblocks corresponds to almost the total number of subblocks, and also needs to be replaced. Therefore, this method is difficult to implement and takes a long time to process. Therefore, in the present invention, the performance is similar to the method of sorting the sum of absolute values of all subblocks in size order, and it is relatively simple to implement the following method. That is, the embodiment will be described in detail with reference to the flowchart of FIG. 3. The absolute sum of subblocks is expressed as an absolute value of M bits as shown in FIG. Therefore, only the target with 2 bit value of 1 is selected in the process (3e), and when it does not correspond to 12% (3k), the remaining blocks are selected for the remaining sub blocks except the sub blocks that are already selected. As the next bit of the MSB is searched again in step 3e, the bit comparison and the selection of the subblock are continued until the number of selected subblocks reaches 12% in step 3g.

In the exemplary embodiment of the present invention, it has been found that the adaptive selector 105 and the adaptive modulator 107 of FIG. 1 are implemented by a computer system corresponding to an image processing apparatus, and can be sufficiently implemented by an ASIC. Is true.

As shown in the computer for an image processing apparatus, the 3D subband analyzer 103 divides each of the remaining 188 subbands into 16 subblocks and calculates energy of each subblock as described above. The energy E of each sub block is calculated by the following equation ②.

Of the detected energy (E), represented by M bits and stored in the computer internal memory (RAM). Since only a predetermined number (22.7) sub-blocks having a large energy value must be adaptively selected from among the energy values of each sub-block indicated by the detected M bits, the CPU inside each computer stores each sub-block energy E stored in the memory. Is processed by the flow chart as shown in FIG. 4, and 22.7 pieces are selected to be transmitted starting from a large value of energy (E). To do this, in step (3a), a predetermined indication bit (M) for the total number (N) of the target subblocks and the absolute value of the sample value in each band is specified, and in step (3b), the selection subindex (SI) is specified. Initialize the register that holds the value.In step (3c), the target bit index (b) to search for energy also specifies the MSB.In step (3d), the absolute value is added to initialize the register to store the index (n). In step 3e, the MSB is checked from the first subblock. Here, when MSB is 0, in order to check the MSB of the next block in step (3i), the sum value is added to increase the index (n). However, when the MSB is 1 in step (3e), the selected subblock index (SI) is read in step (3f), and it is checked whether the number of charge sums (S # 22.7) is reached in step (3g). That is, one block with MSB 1 is generated. When S1 does not reach S in step (3g), the value of n is increased in step (3i) to check next (MSB) by increasing the value of S1 in step (3h). In step (3i), the value of n is increased to check whether the target value of n to be examined is checked in step (3j). When N is not reached here, the process is restarted from the above (3e). However, in step (3g), when S1 does not reach S and reaches the target N to be searched, in step (3k), the bit to search for the MSB is moved as shown in FIG. 6 to check the next bit into the MSB. b) is reduced to check execution as above (3e)-(3j). In the process (3g), when S1 reaches S, all 22.7 pieces, which are 12%, are selected. When the adaptive selection is completed, the modulator adaptively modulates and converts the data into a serial form so that the scrambling unit 109 transmits one channel. The multiplexing modulator 111 further sub-blocks the video signal according to the selection of the selection stage ST. Add selection information and modulation parameters and transmit digital audio and auxiliary data.

As described above, four subbands having high energy are selected and transmitted to reconstruct a better image, and each of the four subbands of each of the YIQs plays a large role in reconstructing the image. In order to separate the digital bit and the information corresponding to one analog value, DPCM is used to show overall good performance, especially 188 except for 4 low frequency band subbands which are always transmitted among 192 subbands of Y component. The high-energy components are adaptively selected for the subbands, but the MSB search method is used to simplify the implementation, and the optimal adaptive selection method reduces the subblock size from 32 to 25.20, thereby improving performance.

Claims (4)

(delete) (delete) (delete) In the data selection circuit including a transmission matrix converter 101, a three-dimensional subband analyzer 103, a scrambling unit 109, and a multiplexing modulator 111, the three-dimensional subband analyzer Adaptively selecting signals of the low frequency band and other bands that are important parts of a given image among the divided n subbands analyzed from 103), and for components of the band except the adaptively selected low frequency band. An adaptive selector 105 for dividing the sub-block into sub-blocks for transmission to a channel having a bandwidth of 6 MHZ and then sequentially selecting blocks having high energy; Adaptively modulates a signal of a low frequency band transmitted directly from the adaptation selector 105 to the scrambling unit 109, and is re-divided for the components of the band except for the directly transmitted low frequency band And an adaptive modulator (107) for adaptively modulating an energy signal and transmitting the modulated energy signal to the scrambling unit (109).