KR100962506B1 - Data Compressing and Decompressing Method and Apparatus, Liquid Crystal Display and Driving Method using the same - Google Patents

Abstract

The present invention relates to a data compression method for compressing and decompressing data.

The data compression method executes a VQ-BTC algorithm on input data to generate first encoded data including a bitmap and a first vector representative value, and encodes a flag bit for identifying the first encoded data. Adding first coded data added to the data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. Generating; Correcting the second vector representative values using the average distance and the correction position; Adding a flag bit for identifying the second encoded data to the second encoded data; And comparing the deviations between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result.

Description

Data Compression and Decompression Method and Apparatus and Liquid Crystal Display and Driving Method Using the Same

The present invention relates to a data compression method and apparatus for compressing and decompressing data, and a method and apparatus for restoring the same. In addition, the present invention provides an over-driving control (ODC) scheme of a liquid crystal display device to convert a data voltage to be applied to liquid crystal cells after compressing and restoring data to be stored in a memory using the data compression and restoration method and apparatus. And a driving method thereof.

As shown in Equations 1 and 2, the liquid crystal display has a disadvantage in that the response speed is slow due to the inherent viscosity and elasticity of the liquid crystal.

(gamma)는 액정분자의 회전점도(rotational viscosity)를 각각 의미한다. Where τr is the rising time when voltage is applied to the liquid crystal, Va is the applied voltage, V _F is the Freederick Transition Voltage at which the liquid crystal molecules start the tilt motion, and d is The cell gap of the liquid crystal cell,

(gamma) means rotational viscosity of liquid crystal molecules, respectively.

Here, τf denotes a falling time during which the liquid crystal is restored to its original position by the elastic restoring force after the voltage applied to the liquid crystal is turned off, and K denotes an elastic modulus inherent to the liquid crystal.

The liquid crystal response speed of TN mode (Twisted Nematic mode), which has been the most commonly used liquid crystal display device, may vary depending on the physical properties of the liquid crystal material and the cell gap. 30 ms. The response speed of the liquid crystal is longer than one frame period (NTSC: 16.67ms). For this reason, as shown in FIG. 1, since the voltage charged in the liquid crystal cell reaches the next frame, motion blur occurs in the video.

Referring to FIG. 1, when the data VD changes from one level to another due to the slow response speed of the liquid crystal, the corresponding display luminance BL may not reach the desired luminance and thus may not display the desired color and luminance. do. As a result, the liquid crystal display exhibits a motion blur in the video and thus the image quality is deteriorated.

In order to solve the slow response speed of the liquid crystal display device, U.S. Pat. Over driving control (ODC) has been proposed. The overdrive method modulates data on the same principle as in FIG. 2.

을 크게 하게 된다. 따라서, 과구동 방법으로 구동되는 액정표시장치는 액정의 늦은 응답속도를 데이터값의 변조로 보상하여 동영상에서 모션 블러를 줄일 수 있다. Referring to FIG. 2, the overdrive method modulates the input data voltage VD into a preset voltage MVD of modulated data and applies the modulated data voltage MVD to a liquid crystal cell to obtain a desired luminance MBL. do. The overdrive method is expressed by Equation 1 based on the change of data so as to obtain a desired luminance (MBL) within one frame period.

To make it larger. Accordingly, the liquid crystal display device driven by the overdrive method may reduce motion blur in the video by compensating for the late response speed of the liquid crystal by modulation of the data value.

This overdrive method compares the data between the previous frame and the current frame and if there is a change between the data, modulates the data of the current frame with preset modulation data.

3 is a block diagram schematically illustrating an overdrive circuit.

Referring to FIG. 3, the overdrive circuit includes first and second frame memories 33a and 33b for storing data from the data input bus 32 and a lookup table 34 for modulating the data. .

The first and second frame memories 33a and 33b alternately store data in units of frames in accordance with the pixel clock, alternately output the stored data, and output the previous frame data, that is, the n-1 th frame data, to the lookup table 34. (Fn-1) is supplied.

The lookup table 34 stores n-th frame data Fn from the n-th frame data Fn and the first and second frame memories 33a and 33b as shown in Table 1 below. Is selected as an address to modulate the data. The lookup table 34 includes a read only memory (ROM) and a memory control circuit.

division 0 One 2 3 4 5 6 7 8 9 10 11 12 13 14 15 0 0 2 3 4 5 6 7 9 10 12 13 14 15 15 15 15 One 0 One 3 4 5 6 7 8 10 12 13 14 15 15 15 15 2 0 0 2 4 5 6 7 8 10 12 13 14 15 15 15 15 3 0 0 One 3 5 6 7 8 10 11 13 14 15 15 15 15 4 0 0 One 3 4 6 7 8 9 11 12 13 14 15 15 15 5 0 0 One 2 3 5 7 8 9 11 12 13 14 15 15 15 6 0 0 One 2 3 4 6 8 9 10 12 13 14 15 15 15 7 0 0 One 2 3 4 5 7 9 10 11 13 14 15 15 15 8 0 0 One 2 3 4 5 6 8 10 11 12 14 15 15 15 9 0 0 One 2 3 4 5 6 7 9 11 12 13 14 15 15 10 0 0 One 2 3 4 5 6 7 8 10 12 13 14 15 15 11 0 0 One 2 3 4 5 6 7 8 9 11 13 14 15 15 12 0 0 One 2 3 4 5 6 7 8 9 10 12 14 15 15 13 0 0 One 2 3 3 4 5 6 7 8 10 11 13 15 15 14 0 0 One 2 3 3 4 5 6 7 8 9 11 12 14 15 15 0 0 0 One 2 3 3 4 5 6 7 8 9 11 13 15

In Table 1, the leftmost column is data of the previous frame Fn-1, and the uppermost row is data of the current frame Fn.

During the nth frame period, the nth frame data Fn is stored in the first frame memory 33a and supplied to the lookup table 34 at the same pixel clock as indicated by the solid line. At the same time, the second frame memory 33b supplies the n−1 th frame data Fn−1 to the lookup table 34 during the n th frame period.

During the n + 1 th frame period, the current n + 1 th frame data Fn + 1 is stored in the second frame memory 33b and supplied to the lookup table 34 at the same pixel clock as indicated by the dotted line. do. At the same time, the first frame memory 33a supplies the nth frame data Fn to the lookup table 34 during the n + 1th frame period.

To reduce the cost of these overdrive circuits, the memory capacity must be reduced. Recently, in order to reduce memory capacity, a method of compressing data stored in a memory has been proposed. Block compression coding (BTC) is well known as such a compression method. The existing 1/3 compressor half BTC algorithm quantizes two levels of data of non-overlapping blocks. When the existing BTC algorithm is applied to a color image, a bitmap block having a predetermined size representing a level for each RGB data and two representative values are required. In general, since each representative value is expressed by the average and the variance, it is possible to achieve a 1/3 compression ratio and further compress data at a 1/4 compression ratio as shown in FIG. 4.

4 shows an example of BTC encoding for encoding input data in units of 4x4 bitmap blocks.

Referring to FIG. 4, the BTC encoding method uses 4 × 4 × 3 × 8 = 384 bits of RGB input data included in a 4 × 4 block to 3 × 4 × 4 = 48 bits of RGB bitmap data (Bitmapr, Bitmapg, Bitmapb) and 2 × 3 × 8 = 48 bits.

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 41, NO. 9 “A method of block truncation coding for color image compression by Takio Kurita and Nobuyuki Otsu” and IEEE TRANSACTIONS ON COMMUNICATIONS VOL. COM-28, NO. 1 The vector quantization BTC (VQ-BTC) compression algorithm proposed by JANUARY 1980, "An Algorithm for Vector Quantizer Design by Yoseph Linde, Andres Buzo and Robert M. Gray", has been developed in order to compress color images more efficiently. Compression effects corresponding to two bitmap blocks can be obtained by encoding one bitmap block and two RGB data as shown in FIG.

FIG. 5 is a diagram illustrating an encoding method of a VQ-BTC compression algorithm in which data values of an original image represented by eight black points are quantized with vector components represented by two white points.

Referring to FIG. 5, in the encoding method of the VQ-BTC compression algorithm, 4 × 4 = 16 bits quantized 4 × 4 × 3 × 8 = 384 bits of RGB input data included in a 4 × 4 block with a vector component of a white point. Since the bitmap block and two representative values 2 × 3 (for RGB) × 8 (for 1 BYTE) = 48 bits are encoded, the input data can be compressed with a compression ratio of 1/6.

However, in the VQ-BTC compression algorithm, if the block size is increased in the longitudinal direction from 2x4 to 2x8, or if the block size is left as it is and the bits allocated to the RGB vector are reduced from 48 bits to 24 bits, the image quality is severely deteriorated. The data error greatly increases during modulation. In other words, the problem of the existing VQ-BTC algorithm is to reduce the number of bits allocated to the representative value of the block to achieve the compression rate, or to increase the block size to try to represent a wide area with two vectors. Image quality deteriorates. If the number of bits assigned to the representative value is reduced, the depth of the image is lowered, which reduces the color expression power, and causes blocking artifacts between neighboring blocks. In addition, when a block of a large area is represented by two vectors, the detailed texture of the image may not be expressed, and blur may be observed in the display image or the edge information may be lost, resulting in degradation at the boundary of the object.

Accordingly, an object of the present invention is to provide a method and apparatus for compressing and restoring data that can reduce the color rendering power and blur of a display image and prevent data loss of edge information. .

Another object of the present invention is to modulate the data voltage to be applied to the liquid crystal cells by the overdrive (ODC) method of the liquid crystal display after compressing and restoring data to be stored in the memory by using the data compression and restoration method and apparatus. A liquid crystal display device and a driving method thereof.

In order to achieve the above object, a data compression method according to an embodiment of the present invention executes a VQ-BTC algorithm on input data to generate first encoded data including a bitmap and first vector representative values and to generate the first encoded data. Adding first coded data to add flag bits for identifying coded data to the first coded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. Generating; Correcting the second vector representative values using the average distance and the correction position; Adding a flag bit for identifying the second encoded data to the second encoded data; And comparing the deviations between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result.

The data compression method further includes dividing a block of input data into smaller subblocks.

The correcting of the second vector representative values may include expressing the second vector representative value and the plurality of vector values existing in the periphery of each of the sub-blocks in an HSI image coordinate system and along the I axis of the HSI image coordinate system. Inversely transforming the image into an RGB image coordinate system; Mapping each of the vector positions represented in the inverse transformed RGB image coordinate system into a codeword; And correcting the second vector representative value by selecting the codeword according to the following mean error equation (MSE).

Where (Rr, Gr, Br) is the second vector representative value of the nth pixel, (Rn, Gn, Bn) is the actual pixel value of the nth pixel, (Rc, Gc, Bc) is the correction position, and 'm' represents the number of vector values in each subblock.

The correcting of the second vector representative values may include expressing the second vector representative value and the plurality of vector values existing in the periphery of each of the sub-blocks in an HSI image coordinate system and along the I axis of the HSI image coordinate system. Inversely transforming the image into an RGB image coordinate system; Mapping each of the vector positions represented in the inverse transformed RGB image coordinate system into a codeword; And calculating each of the following R _code , G _code , and B _code , and selecting the codeword based thereon to correct the second vector representative value.

Here, R ₁ , R ₂ , R ₃ and R ₄ are bit values of the sub block in R data, G ₁ , G ₂ , G ₃ and G ₄ are bit values of the sub block in G data, B ₁ , B ₂ , B ₃ and B ₄ are bit values of the sub block in B data, m is the number of the vector values in each of the sub blocks, R _H is the upper representative value of the second vector representative value in R, R _L is a lower representative value of the second vector representative value.

A data reconstruction method according to an embodiment of the present invention includes a bitmap, vector representative values consisting of upper bits only, an average distance between vector representative values and surrounding vector values, and a correction position of each of the vector values. Receiving encoded data; And decoding the encoded data by using the following equation.

Where (Rd, Gd, Bd) is the decoded pixel value, (Rr, Gr, Br) is the vector representative value, 'L' is the average distance, and (Rc, Gc, Bc) is the correction position, respectively. Indicates.

A data compression apparatus according to an embodiment of the present invention executes a VQ-BTC algorithm on input data to generate first encoded data including a bitmap and a first vector representative value and a flag for identifying the first encoded data. A first encoder for adding first encoded data for adding bits to the first encoded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. A second encoder which generates the second vector representative values by using the average distance and the correction position, and adds a flag bit to the second encoded data to identify the second encoded data; And a selector for comparing the deviation between the vector representatives with a predetermined threshold and selecting the outputs of the first and second encoders according to the comparison result.

An apparatus for restoring data according to an exemplary embodiment of the present invention includes a bitmap, vector representative values consisting of upper bits only, an average distance between vector representative values and vector values around them, and a correction position of each of the vector values. A decoder for decoding coded data is provided.

A liquid crystal display according to an embodiment of the present invention executes a VQ-BTC algorithm on input data to generate first encoded data including a bitmap and a first vector representative value, and a flag for identifying the first encoded data. A first encoder for adding first encoded data for adding bits to the first encoded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. A second encoder that generates the second vector representative values by using the average distance and the correction position, and adds a flag bit for identifying the second encoded data to the second encoded data; A selector for comparing the deviation between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result; And decoding the selected encoded data to generate previous frame data, selecting a modulation value listed in a lookup table using the previous frame data and the current frame data as addresses, and modulating the input data to display the modulated data on the LCD panel. A driving circuit of a liquid crystal display panel to display is provided.

A method of driving a liquid crystal display according to an embodiment of the present invention executes a VQ-BTC algorithm on input data to generate first encoded data including a bitmap and a first vector representative value and to identify the first encoded data. Adding first encoded data that adds a flag bit to the first encoded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. Generating; Correcting the second vector representative values using the average distance and the correction position; Adding a flag bit for identifying the second encoded data to the second encoded data; Comparing the deviations between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result; And decoding the selected encoded data to generate previous frame data, selecting a modulation value listed in a lookup table using the previous frame data and the current frame data as addresses, and modulating the input data to display the modulated data on the LCD panel. Display.

The method and apparatus for compressing and decompressing data according to an embodiment of the present invention executes the VQ-BTC algorithm to perform a bitmap, vector representative values consisting of only upper bits, average distance between vector representative values and vector values in the vicinity thereof. A correction position of each of the vector values may be calculated and the vector representative values may be corrected based on the average distance and the correction position to reduce color deterioration and blur of a display image and to prevent data loss of edge information.

In addition, the liquid crystal display device and the driving method thereof according to an embodiment of the present invention, the data voltage to be applied to the liquid crystal cells after compressing and restoring data to be stored in the memory by using the data compression and decompression method and apparatus. By overmodulating the device, the response characteristics of the liquid crystal can be improved, and the display quality can be improved by preventing color degradation and blur and edge data loss caused by data compression.

Other objects and advantages of the present invention in addition to the above object will become apparent from the description of the preferred embodiment of the present invention with reference to the accompanying drawings.

Hereinafter, exemplary embodiments of the present invention will be described with reference to FIGS. 6 to 25.

Referring to FIG. 6, a liquid crystal display according to an exemplary embodiment of the present invention includes a liquid crystal display panel 100, an overdrive processing circuit 10, a timing controller 11, a data driving circuit 12, and a gate driving circuit ( 13).

In the liquid crystal display panel 100, a liquid crystal layer is formed between two glass substrates. The liquid crystal cells of the liquid crystal display panel 100 are arranged in a matrix by a cross structure of the data lines D1 to Dm and the gate lines 16.

The lower glass substrate of the liquid crystal display panel 100 is connected to the data lines D1 to Dm, the gate lines G1 to Gn, a thin film transistor (TFT), and a TFT to be connected to the pixel electrodes 1. Liquid crystal cells Clc, a storage capacitor Cst, and the like, which are driven by an electric field between the common electrodes 2, are formed.

The black matrix, the color filter, and the common electrode 2 are formed on the upper glass substrate of the liquid crystal display panel 100.

The common electrode 2 is formed on the upper glass substrate in a vertical electric field driving method such as twisted nematic (TN) mode and vertical alignment (VA) mode, and has an in plane switching (IPS) mode and a fringe field switching (FFS) mode. In the same horizontal electric field driving method, the pixel electrode 1 is formed on the lower glass substrate. Polarizing plates having optical axes orthogonal to each other are attached to the upper glass substrate and the lower glass substrate of the liquid crystal display panel 100, and an alignment layer for setting a pre-tilt angle of the liquid crystal is formed at an interface in contact with the liquid crystal.

The overdrive processing circuit 10 compresses and decompresses previous frame data [Fn-1 (RGB)] using a compression and decompression algorithm described below, and a lookup table including modulation data satisfying Equations 3 to 5 below. The current frame data [Fn (RGB)] is compared with the previous frame data [Fn-1 (RGB)] by using Fn (RGB) as the result of comparison. MRGB).

Fn (RGB) <Fn-1 (RGB) ---> Fn (MRGB) <Fn (RGB)

Fn (RGB) = Fn-1 (RGB) ---> Fn (MRGB) = Fn (RGB)

Fn (RGB)> Fn-1 (RGB) ---> Fn (MRGB)> Fn (RGB)

As can be seen from equations (3) to (5), the lookup table of the overdrive processing circuit 10 has a current frame if the current frame data [Fn (RGB)] is larger than the previous frame data [Fn-1 (RGB)]. Modulates the current frame data [Fn (RGB)] to a value MRGB greater than the data [Fn (RGB)], while the current frame data [Fn (RGB)] is greater than the previous frame data [Fn-1 (RGB)]. Smaller] modulates the current frame data [Fn-1 (RGB)] with a smaller value (MRGB) than the current frame [Fn (RGB)]. The lookup table of the overdrive processing circuit 10 outputs the same value MRGB as the current frame data Fn (RGB) if the pixel data value of the same pixel is the same in the previous frame and the current frame.

The timing controller 11 includes a data timing control signal for controlling the operation timing of the data driving circuit 12 using timing signals such as vertical and horizontal synchronization signals, data enable signal DE, and dot clock CLK. The gate timing control signal for controlling the operation timing of the gate driving circuit 13 is generated. In addition, the timing controller 11 transmits the overdrive modulated digital video data MRGB to the data driving circuit 12.

The data driving circuit 12 converts the overdrive-modulated digital video data MRGB from the timing controller 11 into the positive / negative gamma compensation voltage in response to the data timing control signal to convert the analog positive / negative polarity. A data voltage is generated and the data voltage is supplied to the data lines D1 to Dm.

The gate driving circuit 13 sequentially shifts the gate pulse in response to the gate timing control signal to supply the gate pulse, that is, the scan pulse to the gate lines G1 to Gn.

7 is a block diagram showing the overdrive processing circuit 10 in detail. 8 is a diagram illustrating in detail a control process of the encoder 20 of FIG. 7.

7 and 8, the overdrive processing circuit 10 includes an encoder 20, a decoder 24, and a lookup table 25.

The encoder 20 comprises first and second encoders 21, 22.

The first encoder 21 encodes the digital video data RGB by executing the VQ-BTC algorithm on a 2x8 or 4x4 block basis and outputs the encoded data packet as shown in FIG. 9.

The second encoder 22 encodes digital video data (RGB) with a VQ-BTC (hereinafter, referred to as "P_VQ-BTC") compression algorithm proposed in the embodiment of the present invention.

In P_VQ-BTC, the representative value is assigned to the upper vector representative value in each of RGB by the most significant bit (MSB) i (i is 4 or 5) bits, and the lower vector is assigned the most significant bit (MSB) j (j is 5 or 4) bits. do. If the most significant bits (MSB) 4 bits are allocated to the upper vector representative value, the most significant bits (MSB) 5 bits are allocated to the lower vector, and if the most significant bits (MSB) 5 bits are allocated to the upper vector representative value, the most significant bits ( MSB) 4 bits are allocated. If the most significant bits (MSB) 4 bits are allocated to the upper vector representative value and the most significant bits (MSB) 5 bits are allocated to the lower vector, 3 bits each of the upper vector and 4 bits of the lower vector are compared with the conventional VQ-BTC algorithm. Three savings result in a total of 21 bits in the encoded data packet compared to the existing VQ-BTC compression algorithm. This margin is allocated to the average distance and location information.

) 또는 맨하탄 거리(Manhattan Distance :

)로 계산될 수 있다. 여기서, 평균 거리는 비트맵을 사용하여 서브 블록 안의 픽셀 벡터가 하위 그룹으로 구분된다면 하위 대표 벡터[R₀ G₀ B₀]^T와의 차로 계산되고 서브 블록 안의 픽셀 벡터가 상위 그룹으로 구분된다면 상위 대표 벡터[R₁ G₁ B₁]^T와의 차로 계산된다. The P_VQ-BTC algorithm executes the VQ-BTC algorithm based on a 2x8 block to determine the average distance between 16 vector values in the encoded data and the vector representative value obtained as the average value. (S1) The average distance is the Euclidean distance of all the pixel vectors [R _n G _n B _n ] ^T and the representative vector [R _K G _K B _K ] ^{T in} the subblock, as shown in Equation 6.

) Or Manhattan Distance (

Can be calculated as Here, the average distance is calculated by the difference from the lower representative vector [R ₀ G ₀ B ₀ ] ^T if the pixel vectors in the subblock are divided into subgroups using the bitmap, and the upper representative vector if the pixel vectors in the subblock are divided into higher groups. Calculated as the difference with [R ₁ G ₁ B ₁ ] ^T.

Here, m represents the number of the vector values in each of the subblocks.

Among the 16 vectors, the vector belonging to group 1 is used to calculate the distance from the upper vector, and the vector belonging to group 2 is used to determine the distance from the lower vector. In P_VQ-BTC, the average distance is allocated 4 bits and stored as one of the values 0x0004-0x003c, that is, 4 to 60. P_VQ-BTC divides the block of input data into 2 × 2 subblocks (S2), and corrects the vector representative value by using the average distance obtained in units of each sub-block and 15 possible positional information corrections. (S3) A detailed description of a method for finding a position that may be corrected will be given later. P_VQ-BTC encodes the correction position in the form of a 4-bit codeword for each sub-block, and allocates four 16 bits to the encoded data stream, each of four bits. Therefore, the data stream encoded in P_VQ-BTC includes 4 bits allocated as the average distance and 16 bits allocated as the correction position.

If the deviation of the vector representative values in the data encoded by the first encoder 21 with the VQ-BTC compression algorithm is smaller than two or more thresholds, the P_VQ-BTC does not go through the steps S1 to S3 and then the first encoder 21. (S0) The encoder 20 selects the outputs of the first and second encoders 21 and 22 by comparing the deviations of the vector representative values and the threshold values. It includes a selector. Here, the threshold value is set to 8 in the experiment described later as a positive integer.

The encoder 20 allocates 1 bit to indicate the VQ-BTC or P_VQ-BTC compression algorithm selected by the step S0. In order to allocate this 1 bit, the last bit of any one of the lower vectors of R, G, and B data may be used as a flag bit indicating a compression algorithm. In the experiment described later, the flag 1 bit was used as the remaining bit by deleting the last 1 bit of the R data.

The above-described P_VQ-BTC algorithm compresses an image by dividing it into 2 × 8 blocks to prevent blocking artifacts, and further divides the block into four 2 × 2 sub-block units to express the boundary value more precisely. The VQ-BTC algorithm codes a 16-bit bitmap and six 8-bit representative values to compress a 2x8 block. In contrast, the P_VQ-BTC algorithm uses a 16-bit bitmap, three 5-bit vector representations, three 4-bit vector representations, four bits of distance information, and a 16-bit subblock unit to compress a 2x8 block. Code position information and 1 bit flag. The encoding process of the above-described P_VQ-BTC is summarized as follows.

(1) The existing BTC algorithm is executed based on 2 × 8 blocks. At this time, the representative value is allocated by storing 5 bits as MSB in the upper vector and 4 bits as MSB in the lower vector.

(2) The average distance between the 16 vector values and the vector representative value is calculated.

(3) A 2x8 block is divided into 2x2 subblocks.

(4) The vector representative value is corrected by the above-described correction position determination methods based on the average distance and correction position calculated in (2) in each sub-block unit.

(5) The correction position is encoded in the form of a 4-bit codeword for each sub-block.

(6) If the average distance is smaller than the threshold value, the data is compressed in units of 2 × 8 blocks using the existing VQ-BTC algorithm without going through steps (3) to (5).

The decoder 24 decodes the data compressed by the encoder 20 and supplies it to the lookup table 25. The decoder 24 reads the flag bits to determine whether the encoding method of the input encoded data is VQ-BTC or P_VQ-BTC to select an appropriate decoding method.

The lookup table 25 lists overdrive modulation values satisfying Equations 3 to 5.

9 is a data packet protected by the first encoder 21.

Referring to FIG. 9, one data packet encoded by the first encoder 21 includes a flag bit of 1 bit, a 2x8 or 4x4 bitmap of 16 bits, and (2x8) -1 = 15 bits. Two vector representatives of R data assigned to R, two vector representatives of G data assigned to 2x8 = 16 bits, and two vector representative values of B data assigned to 2x8 = 16 bits. do. Thus, one data packet encoded by the first encoder 21 includes 64 bits.

10 is a data packet encoded by the second encoder 22.

Referring to FIG. 10, one data packet encoded by the second encoder 22 includes a flag bit of 1 bit, a 2x8 or 4x4 bitmap of 16 bits, 4 (higher vector MSB) +5 (lower vector). MSB) = two vector representatives of R data allocated with 9 bits, 4 + 5 = two vector representatives of G data allocated with 9 bits, two vectors of B data allocated with 4 + 5 = 9 bits The representative value, average distance information of 4 bits, and correction position information of 4 (RGB 2x2 sub-block) x 4 (correction position codeword) = 16 bits are included. Therefore, one data packet encoded by the second encoder 22 includes 64 bits.

FIG. 11 is one data packet generated by the existing VQ-BTC compression algorithm as shown in FIG. 5.

Referring to FIG. 11, one data packet generated by the existing VQ-BTC compression algorithm includes a bitmap of 16 bits and representative values of 16 bits allocated to each of RGB. Therefore, one data packet generated by the existing VQ-BTC compression algorithm includes 64 bits in total and does not include flags, average distance information, and corrected position information.

In encoding the vector representative value, the second encoder 22 does not use 8 bits in each of RGB, but uses only upper 4 bits or upper 5 bits. Therefore, the decoder 24 adds "0" to the lower 3 bits or 4 bits as shown in FIG. 12B when decoding the vector representative value when the original vector representative value is as shown in FIG. 12, or "0111" or "as shown in FIG. 12C. 011 "is added to the lower 3 bits or 4 bits. Among the decoding methods of the vector representative value, in order to minimize image quality deterioration due to cutting of the lower bits, a method of adding the lower bits as shown in FIG. 12C is preferable.

In FIG. 8, the correction position determining method processed in step S3 of the second encoder 22 will be described in detail as follows.

First embodiment of the correction position determination method

In order to properly correct the sub-blocks, the correction position must be determined. Equation 7 shows the relationship between the vector representative value and the actual pixel value.

,

,

)는 n 번째 픽셀에서 실제 픽셀값과 벡터 대표값 간의 오차값이다. Here, (Rr, Gr, Br) is the vector representative value of the nth pixel, and (Rn, Gn, Bn) is the actual pixel value of the nth pixel. And (

,

,

) Is an error value between the actual pixel value and the vector representative value at the n th pixel.

,

,

)는 수학식 8과 같이 모델링될 수 있다.When 'L' is the average distance between the vector representative value and the 16 vector values, and (Rc, Gc, Bc) is the correction position,

,

,

) May be modeled as in Equation 8.

Although all pixel values around the vector representative value can be corrected, the data compression and decompression method according to an embodiment of the present invention determines the corrected position in the HSI (Hue, Saturation, Intensity) image coordinate system to determine a more effective correction position. do. The HSI image coordinate system is a coordinate system that can express colors focusing on human vision system (HVS), and is an image coordinate system that is attracting the most attention in recent image processing. The conversion method of the HSI image coordinate system in the RGB image coordinate system is shown in FIG. 13.

Referring to the RGB image coordinate system as shown in FIG. 13A, an imaginary line (I axis of the HSI image coordinate system) connecting brightness (0, 0, 0) to (255, 255, 255) is aligned with the B axis. If moved, it is as shown in Fig. 13 (b). FIG. 13B may be converted to a similar HSI image coordinate system as shown in FIG. 13C. In this HSI image coordinate system, a vector representative value and 16 vector values around it can be expressed. For conversion of RGB image coordinate system and HSI image coordinate system, see CH2898-5 / 90/0000/0791 $ 01.00 © 1990 IEE “Fundamentals of True-Color Image Processing by RS Ledley, M. Buas, and TJ Golab”, CH2998- 3/91 / 10000-0722 $ 01.00 © 1991 "IMAGE PROCESSING USING THE HSI COLOR SPACE by Eric Welch, Robert Moorhead, and JK Owens" disclosed in the IEE, and so will be omitted in detail.

As shown in the left figure of FIG. 14, the HSI image coordinate system is divided into seven sections according to the intensity (I-axis) difference, and when the points corresponding to the respective sections are found, the positions of the points are determined. 14 is a diagram showing the position of each vector (black point) in the HSI image coordinate system.

In other words, when the HSI image coordinate system is inversely transformed into an RGB image coordinate system, the coordinates of the black dots in the inversely transformed RGB image coordinate system as shown in FIG. 15 are positions of vector values at which the vector representative value can be corrected. Each of the black dots may be converted into a 4-bit codeword as shown in FIG. 15 and may be created as a table as shown in Table 2 below. Such a table may be implemented as a lookup table embedded in the second encoder 22. P_VQ-BTC performs encoding based on the table in Table 2 to correct the vector representative value in each subblock. Accordingly, the second encoder 22 can simply correct the vector representative value by searching the table of Table 2 for a position that can be corrected when the vector representative value is corrected.

In Table 2 above, P_VQ-BTC selects a codeword (c) that satisfies a minimum mean square error (MSE) as shown in Equation 9 below to select a correction position having the smallest error.

Where (Rr, Gr, Br) is the vector representative value of the nth pixel, (Rn, Gn, Bn) is the actual pixel value of the nth pixel, (Rc, Gc, Bc) is the correction position, and 'm' is Represents the number of vector values in each subblock.

Second Embodiment of Correction Positioning Method

In the method of determining a correction position according to the second embodiment of the present invention, the correction position can be quickly selected by using the following algorithm without performing a process for selecting all the correction positions as shown in Equation (9).

(1) Since one codeword is applied to the 2x2 subblock regardless of the upper representative value and the lower representative value, the codeword can be generated if the correlation between each pixel and the vector representative value is determined. Since each pixel value and vector representative value are known, the pixel values R _o , G _o , and B _o are equal to the vector representative values R _r , as shown in Equation 10 below. _r G, _r B) is greater than "+1", the pixel values (r _o, G _o, B _o) and a representative value vector (if r _{r, r} G, _r B) is similar to "0", the pixel values ( If R _o , G _o , B _o ) is less than the vector representative values R _r , G _r , and B _r , select '-1' to add the selected values from each 2 × 2 sub-block of each RGB. Here, similarity means that the difference between the pixel value and the vector representative value is less than half the average distance.

In Equation 10, (R _o , G _o , B _o ) are original pixel values of the input image and (R _r , G _r , B _r ) are vector representative values of respective RGB.

In order to correct the vector correction value in the 2x2 subblock of R, the R _code for selecting the index of Table 2 has a value from +4 to -4. Similarly, to correct the vector correction value in the 2 × 2 subblock of G, the G _code for selecting the index in Table 2 has a value from +4 to -4, and the vector correction in the 2 × 2 subblock of B. To correct the value, the B _code for selecting the index in Table 2 has a value from +4 to -4.

R _code , G _code and B _code added in 2x2 subblock, respectively, if the value is 4,3,2, '1' in the codewords in Table 2, R _code , G _code and B added in 2x2 subblock, respectively _code , if the value is 1, 0, -1, R _code , G _code , B _code added in '0' and 2x2 subblocks in the codeword of table 2, and table if the value is -2, -3, -4 By selecting '-1' from the codeword of 2, the codeword is searched in the order of R->G-> B as shown in FIG. 16, and as a result, a preset index value is found in the rectangle. For example, if the R _code added in the 2x2 subblock is '-1', the G _code added in the 2x2 subblock is '1', and the B _code added in the 2x2 subblock is '1', the vector in Table 2 The index indicating the correction position of the representative value is '5', and therefore, the correction position of the vector representative value is determined as (-1, 0, -1) in Table 2 and FIG. This embodiment has the advantage of fast and simple operation speed by determining the correction position of the vector representative value by three operations.

In the tree type search structure of FIG. 16, when a codeword is not assigned (No) may occur due to structural features. For example, R _code , G _code , and B _code may be determined to be -1, 0, or 0, respectively, but there is no corresponding index. In such a case, the index is determined by replacing similar codewords in the vicinity in the direction of minimizing the error as described below with reference to FIGS. 17 to 19.

Since the codewords in Table 2 do not reflect each vertex of the cube, the center of the line, and the center of the plane as shown in FIGS. 14 and 15, the index when searching the index for selecting a correction position in the same manner as described above. There is a part that cannot be mapped. At this time, all parts not included in the codeword, for example, (1, 0, 0) and the like are all mapped to (0, 0, 0). As a result, the error centered on B was distributed to R, G, and B. Based on the experimental results, the average PSNR (Peak Signal to Noise ratio) and image quality were further improved. The effects on this can also be explained with reference to FIGS. 17 and 18.

Assuming that the vector representative values are points [(0,0,0), (-1,1,0), (-1, -1, -1)] as shown in FIG. 18 is calculated as to which point is closer to any one point. FIG. 18A is a diagram illustrating the positions of the above three points [(0,0,0), (-1,1,0), (-1, -1, -1)] in the RGB image coordinate system. . FIG. 18B shows the positions of intermediate vectors of (0,0,0) and (-1,1,1). FIG. 18C shows the positions of intermediate vectors of (0,0,0) and (-1,1,0).

In FIG. 18, since (-1,1,0) does not exist in the codewords of FIGS. 15 and 2, the points exist at (0,0,0) and (-1,1,0) centers and 3/4 positions. They may be mapped to (-1, 1, 0) as shown in (c) of FIG. 18 and then back to (-1, 1, 1). That is, although the hatched portion in FIG. 18C is close to (0,0,0), an error of vector representative value correction mapped to (-1,1,0) may occur. In order to prevent such an error, it is preferable to correct the tree search structure of FIG. 16 as shown in FIG. . The index (0,0,0) of FIG. 19 is a value having the most similar intensity in the HSI image coordinate system and combinations that may be selected as the NO code in FIG. 16. In addition, in order to more accurately select a codeword of a subblock, Equation 10 is modified to Equation 11 below.

Here, m represents the number of lower vectors in the 2x2 subblock, R _H is an upper representative value of R, and R _L is a lower representative value. The appropriateness of the codeword calculated through Equation 11 may be determined by Equation 12 below.

Here, "aver_dis" is an average distance between the vector representative value and the 16 vectors around it.

After calculating the value of applying the equation (12) for each of the RGB, it is necessary to select a codeword that best matches the combination of the calculated values. In Equation 12, the average distance is multiplied by 2 because there are actually 4 pixel values in the 2 × 2 subblock, and R _code is obtained by adding the difference between each point corresponding to the pixel values and the vector representative value. This is because the _code must be recognized as +1 when the distance is multiplied by four times the average distance.

20 is a diagram illustrating decoding of a vector representative value between an existing VQ-BTC compression algorithm and a data decompression method according to an embodiment of the present invention.

20A illustrates a decoding method of the existing VQ-BTC. In contrast, the vector representative value reconstruction of the data reconstruction method according to the embodiment of the present invention decodes the vector representative value in units of 2 × 2 subblocks as shown in FIG. 20B and as shown in FIG. 20C. The pixel value is corrected by correcting the vector representative value in units of subblocks. 20D is a diagram illustrating a corrected vector of pixel values reconstructed by the data reconstruction method according to the embodiment of the present invention.

A data restoration method according to an embodiment of the present invention is shown in Equation 13 below.

Here, (Rd, Gd, Bd) is the decoded pixel value, (Rr, Gr, Br) is the vector representative value of the pixel, 'L' is the average distance between the vector representative value and the surrounding vector values, (Rc, Gc, Bc) are correction positions. The decoder 24 reads the flag bits of the encoded data to determine the encoding method of the input encoded data. When the encoded data encoded by the VQ-BTC algorithm by the first encoder 21 is input to the decoder 24, the decoder 24 sets the average distance L and the bits of the correction positions Rc, Gc, and Bc to 0. The encoded data is decoded by using the VQ-BTC method.

In the step S0 of FIG. 8, the vector representative values are selected when the VQ-BTC-based compression algorithm and the P_VQ-BTC compression algorithm proposed by the present invention are selected by comparing the deviation of the vector representative values with the threshold value. When the deviation is less than the threshold, the VQ-BTC based compression algorithm is selected. This is because P_VQ-BTC divides the vector groups into 2 × 2 sub-blocks and corrects the vector representative value by the above-described correction positioning method, thereby preventing artifacts between blocks without degrading color expression power even when the vector representative value is large. Because it can. On the other hand, when the variation of the vector representative value is large, the existing VQ-BTC has a low color depth of the image, thereby reducing color expressiveness and generating artifacts between neighboring blocks. On the other hand, in the conventional VQ-BTC, when the variation of the vector representative value is small, almost no artifacts between blocks are observed, and since there is no separate calculation process like the above-described correction positioning method, the P_VQ-BTC compression is simple and fast. Advantageous over the algorithm.

21A to 24E compress the display image after compression and reconstruction of the existing BTC algorithm for each of the four experimental images, the display image after compression and reconstruction of the existing VQ-BTC algorithm, and the P_VQ-BTC (proposed_VQ_BTC) algorithm. This is an experimental result of the display image after reconstruction using the correction position determination method of the first embodiment, the display image after compression using the P_VQ-BTC algorithm, and reconstruction using the correction position determination method of the second embodiment. Table 3 is a quantitative evaluation of PSNR performance for the experimental results of FIGS. 21A-24E.

Legacy BTC (1/3 Compression) Conventional VQ-BTC (1/6 Compression) P_VQ-BTC (1/6 Compression), Correction Positioning Method of First Embodiment P_VQ-BTC (1/6 Compression), Correction Positioning Method of Second Embodiment Parrot
(768 x 512)
36.21 33.96 34.62 34.24 Korea014
(1000x668)
31.93 30.24 30.89 30.66 Cafe (1280x768)
28.73 26.54 27.56 27.47 Fruit-XW
(1280x768) 33.76 32.15 32.64 32.34

As can be seen from Table 3, it can be seen that the PSNR evaluation result of P_VQ-BTC proposed in the present invention is high in any experimental image, and it was confirmed that color expression power and blur were improved even in visual evaluation. For example, the experimental image of FIG. 22 will be described. FIG. 25 is an enlarged experimental image of portions of FIGS. 22B to 22D. 25 (a) is an experimental image of the existing BTC algorithm (1/3 compression), FIG. 25 (b) is an experimental image of the existing VQ-BTC algorithm (1/6 compression), and FIG. 25 (c) is Experimental images to which the complementary positioning method of the first embodiment is applied in P_VQ-BTC are respectively shown. Comparing the images in the box of FIG. 25, it can be seen that blur is improved in FIG. 25C even with the naked eye.

In order to help the understanding of the present invention, the encoding process of the BTC algorithm and the VQ-BTC algorithm will be described as an example.

The BTC algorithm applied a scalar quantizer for each of RGB. In contrast, the VQ-BTC algorithm processed by the first encoder 21 represents each of the RGB as a vector of (R, G, B) and applies a vector quantizer using these vectors.

A method of determining a threshold in the existing BTC algorithm and the VQ-BTC algorithm will be described with reference to the example of FIG. 26.

FIG. 26 is an example RGB block of an experimental image input to the first encoder 21. In the (136,188) part of the image (YCbCr 4: 2: 0), the block is extracted in a 2 × 4 block size.

In the conventional BTC algorithm, an average value of pixel values in a block is determined as a quantization threshold. That is, in the conventional BTC algorithm, different thresholds are assigned to each of RGB. In the conventional BTC algorithm, since different thresholds are allocated to each of RGB, color distortion may occur due to the threshold difference. Since the left-most upper RGB pixel value of the 2x4 block is (229, 210, 200), the image is close to white gradation, but (248, 158, 148) encoded by the BTC algorithm is gradation close to pink. If the YCbCr 4: 2: 0 conversion algorithm is added to the BTC algorithm, the color distortion looks more severe.

768 × 512 parrot as shown in Fig. 26 The process of encoding the block extracted with the 2 × 4 block size in the (136,188) portion of the image (YCbCr 4: 2: 0) by the VQ-BTC algorithm is as follows.

FIG. 27 shows an example of converting pixel values in each block of RGB into vector values by grouping elements at the same position. The VQ-BTC algorithm quantizes the vector values using k-mean clustering. The method can be classified step by step as follows.

(1) Calculate the distance between each vector value. As described above, the Euclidean distance calculation method or the Manhattan distance calculation method may be applied as described above.

(2) Select the longest distance from the calculated distances and determine which vector values are spaced apart from the longest distance.

(3) Based on the calculated two vector values, the remaining vectors are divided by the nearest vector center. As a result, the vector values in each block are divided into two groups, that is, an upper block and a lower block.

(4) The average vector (vector representative value) is calculated using the vector values in each group.

(5) Repeat (3) to (5) again based on the calculated average vector.

This process will be described by way of example with reference to FIGS. 28 to 30.

28 shows the Manhattan distance between the vector values. In FIG. 28, the longest distance is 363, and corresponding vector values are (255, 255, 255) and (150, 131, 121). Based on the calculated vector value, the remaining six vectors are divided into two groups as shown in FIG. 29.

Referring to FIG. 29, vector values are divided into two groups according to distances. The VQ-BTC algorithm calculates the mean of each group using the vector values in each group. If the group to which (255, 255, 255) belongs is Group 1 and the rest of the group is Group 2, the average of Group 1 is (247, 242, 238), and the average of Group 2 is (162, 143, 133). The VQ-BTC algorithm uses the average thus obtained to re-measure the distance between the vectors and the mean vector (vector representative value) and reassemble the group based on this. As a result, the grouping and vector values of the VQ-BTC are as shown in FIG. In Fig. 30, the average vector is a vector representative value of each group. The vector representative value of the group 1 is the upper representative value, and the vector representative value of the group 2 is the lower representative value.

Those skilled in the art will appreciate that various changes and modifications can be made without departing from the technical spirit of the present invention. Therefore, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification but should be defined by the claims.

1 is a waveform diagram showing a change in luminance according to data in a conventional liquid crystal display.

2 is a waveform diagram showing an example of a luminance change caused by data modulation in the overdrive method.

3 is a circuit diagram showing an overdrive circuit.

4 is a diagram illustrating an example of a conventional BTC compression algorithm.

5 is a diagram illustrating an example of a conventional VQ-BTC compression algorithm.

6 illustrates a liquid crystal display according to an exemplary embodiment of the present invention.

FIG. 7 is a block diagram illustrating in detail the overdrive processing circuit shown in FIG. 6.

FIG. 8 is a diagram illustrating the encoder and the processing procedure shown in FIG. 7 step by step.

9 is a diagram showing a data packet encoded by the first encoder 21 with the VQ-BTC compression algorithm.

10 is a data packet encoded by the P_VQ-BTC compression algorithm by the second encoder 23.

11 is a diagram illustrating a data packet generated by a conventional VQ-BTC compression algorithm.

12A to 12C are diagrams illustrating an example of decoding a vector representative value.

FIG. 13 is a view illustrating a conversion process between an RGB image coordinate system and an HSI image coordinate system in the method for determining a corrected position of a vector representative value according to the first embodiment of the present invention.

FIG. 14 is a diagram illustrating vector positions present in HSI image coordinate systems and vector positions existing according to I-axis differences between HSI image coordinate systems.

FIG. 15 is a diagram illustrating an example of mapping vector positions where a vector representative value may be corrected to a 4-bit codeword in an inverse transformed RGB image coordinate system.

16 is a diagram illustrating a tree search in the method for determining the position of correction of the vector representative value according to the second embodiment of the present invention.

FIG. 17 is a diagram showing three arbitrary points at which a vector representative value can be corrected in a cube.

FIG. 18 is a diagram illustrating an error occurring when a correction position existing between three points as shown in FIG. 17 but not present in the codewords of FIGS. 15 and 2 is replaced with a similar correction position.

FIG. 19 is a diagram illustrating a result of substituting (0,0,0) for a correction position that does not exist in the codewords of FIG. 15 and Table 2 in FIG.

20 is a diagram illustrating a decoding method of VQ-BTC and P_VQ-BTC.

21A is the original image of the experimental image "Parrot (resolution: 768 x 512)".

21B is an image showing a display image after compression and reconstruction of the existing BTC algorithm for the experimental image "Parrot (resolution: 768 x 512)".

FIG. 21C is an image showing a display image after being compressed and reconstructed with the existing VQ-BTC algorithm for the experimental image “Parrot (resolution: 768 × 512)”.

FIG. 21D is an image showing a display image after compression with the P_VQ-BTC algorithm for the experimental image "Parrot (resolution: 768 x 512)" and restored using the correction position determination method of the first embodiment.

FIG. 21E is an image showing a display image after compression by the P_VQ-BTC algorithm for the experimental image "Parrot (resolution: 768 x 512)" and restored using the correction position determination method of the second embodiment.

22A is an original image of the experimental image "Korea014 (resolution: 1000 x 668)".

FIG. 22B is an image showing a display image after compression and reconstruction of the existing BTC algorithm for the experimental image “Korea014 (resolution: 1000 × 668)”. FIG.

FIG. 22C is an image showing a display image after being compressed and reconstructed with the existing VQ-BTC algorithm for the experimental image “Korea014 (resolution: 1000 × 668)”.

FIG. 22D is an image showing a display image after compression by the P_VQ-BTC algorithm for the experimental image “Korea014 (resolution: 1000 × 668)” and restored using the correction position determination method of the first embodiment.

FIG. 22E is an image showing a display image after being compressed by the P_VQ-BTC algorithm for the experimental image “Korea014 (resolution: 1000 × 668)” and reconstructed using the correction position determination method of the second embodiment.

23A is an image showing an experimental image "cafe (resolution: 1280 x 768)."

FIG. 23B is an image showing a display image after compression and reconstruction of the existing BTC algorithm for the experimental image “cafe (resolution: 1280 × 768)”.

FIG. 23C is an image showing a display image after being compressed and reconstructed with the existing VQ-BTC algorithm for the experimental image “cafe (resolution: 1280 × 768)”.

FIG. 23D is an image showing a display image after being compressed by the P_VQ-BTC algorithm for the experimental image "cafe (resolution: 1280 x 768)" and restored using the correction position determination method of the first embodiment.

Fig. 23E is an image showing a display image after compression by the P_VQ-BTC algorithm for the experimental image "cafe (resolution: 1280 x 768)" and restored using the correction position determination method of the second embodiment.

24A is the original image of the experimental image "Fruit-XW (resolution: 1280 x 768)".

24B is an image showing a display image after compression and reconstruction of the existing BTC algorithm for the experimental image “Fruit-XW (resolution: 1280 × 768)”.

24C is an image showing a display image after being compressed and reconstructed with the existing VQ-BTC algorithm for the experimental image "Fruit-XW (resolution: 1280 x 768)."

FIG. 24D is an image showing a display image after being compressed with the P_VQ-BTC algorithm for the experimental image "Fruit-XW (resolution: 1280 x 768)" and restored using the correction position determination method of the first embodiment.

FIG. 24E is an image showing a display image after compression by the P_VQ-BTC algorithm for the experimental image "Fruit-XW (resolution: 1280 x 768)" and restored using the correction position determination method of the second embodiment.

FIG. 25 is an enlarged image of a portion of FIGS. 22B to 22D.

Fig. 26 is a view showing 2 × 4 blocks extracted from the experimental image (136,188 part of parrot) and their pixel values.

27 is a diagram illustrating an example in which pixel values in a block are converted to vector values by the VQ-BTC algorithm.

FIG. 28 is a diagram illustrating a Manhattan distance between the vector values of FIG. 27.

FIG. 29 is a diagram illustrating an example grouping based on distances between vector values of FIG. 28.

30 is a diagram illustrating a vector representative value determined by the VQ-BTC algorithm.

Description of the Related Art

10: overdrive processing circuit 21, 22, 23: encoder

24: Decoder 25: Lookup Table

Claims (11)

(a) executing a VQ-BTC algorithm on the input data to generate first encoded data including a bitmap and first vector representative values and assigning a flag bit for identifying the first encoded data to the first encoded data; Adding first encoded data to be added; (b) a second including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values; Generating encoded data; (c) correcting the second vector representative values using the average distance and the correction position; (d) adding flag bits for identifying the second coded data to the second coded data; And (e) comparing the deviations between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result. The method of claim 1, And dividing a block of input data into smaller sub-blocks between (b) and (c). The method of claim 2, In step (c), Expressing the second vector representative value and the plurality of vector values in the periphery of each of the sub-blocks in an HSI image coordinate system, and dividing the vector values along an I axis of the HSI image coordinate system, and then inversely converting the RGB vector coordinate system into an RGB image coordinate system ; Mapping each of the vector positions represented in the inverse transformed RGB image coordinate system into a codeword; And Selecting the codeword according to the following mean error equation (MSE) to correct the second vector representative value, And the second vector representative value is an average value of the vector values existing around the second vector representative value.

Where (Rr, Gr, Br) is the second vector representative value of the nth pixel, (Rn, Gn, Bn) is the actual pixel value of the nth pixel, (Rc, Gc, Bc) is the correction position, and 'm' represents the number of the vector values in each of the subblocks. The method of claim 2, In step (c), Expressing the second vector representative value and the plurality of vector values in the periphery of each of the sub-blocks in an HSI image coordinate system, and dividing the vector values along an I axis of the HSI image coordinate system, and then inversely converting the RGB vector coordinate system into an RGB image coordinate system ; Mapping each of the vector positions represented in the inverse transformed RGB image coordinate system into a codeword; And Calculating each of the following R _code , G _code , and B _code , and selecting the codeword based thereon to correct the second vector representative value, And the second vector representative value is an average value of the vector values existing around the second vector representative value.

Here, R ₁ , R ₂ , R ₃ and R ₄ are bit values of the sub block in R data, G ₁ , G ₂ , G ₃ and G ₄ are bit values of the sub block in G data, B ₁ , B ₂ , B ₃ and B ₄ are bit values of the sub block in B data, m is the number of the vector values in each of the sub blocks, R _H is the upper representative value of the second vector representative value in R, R _L is a lower representative value of the second vector representative value. Receiving encoded data including a bitmap, vector representative values consisting of upper bits only, an average distance between the vector representative values and the surrounding vector values, and a correction position of each of the vector values; And And decoding the encoded data by using the following equation.

Where (Rd, Gd, Bd) is the decoded pixel value, (Rr, Gr, Br) is the vector representative value, 'L' is the average distance, and (Rc, Gc, Bc) is the correction position, respectively. Indicates. Performing a VQ-BTC algorithm on the input data to generate first encoded data comprising a bitmap and a first vector representative value and adding flag bits for identifying the first encoded data to the first encoded data; A first encoder for adding one encoded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. A second encoder which generates the second vector representative values by using the average distance and the correction position, and adds a flag bit to the second encoded data to identify the second encoded data; And And a selector for comparing the deviation between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result. The method of claim 6, The second encoder, The block of input data is divided into smaller sub-blocks, and the second vector representative value and the plurality of vector values existing in the periphery of each of the sub-blocks are represented in an HSI image coordinate system, and the vector value is along an I axis of the HSI image coordinate system. Split them and then inversely transform them into an RGB image coordinate system, After mapping each of the vector positions represented in the inverse transformed RGB image coordinate system to a codeword, Selecting the codeword according to the following mean error equation (MSE) to correct the second vector representative value, And the second vector representative value is an average value of the vector values existing around the second vector representative value.

Where (Rr, Gr, Br) is the second vector representative value of the nth pixel, (Rn, Gn, Bn) is the actual pixel value of the nth pixel, (Rc, Gc, Bc) is the correction position, and 'm' represents the number of vector values in each subblock. The method of claim 6, The second encoder, The block of input data is divided into smaller sub-blocks, and the second vector representative value and the plurality of vector values existing in the periphery of each of the sub-blocks are represented in an HSI image coordinate system, and the vector value is along an I axis of the HSI image coordinate system. Split them and then inversely transform them into an RGB image coordinate system, After mapping each of the vector positions represented in the inverse transformed RGB image coordinate system to a codeword, Calculate the following R _code , G _code , and B _code , respectively, and select the codeword based thereon to correct the second vector representative value, And the second vector representative value is an average value of the vector values existing around the second vector representative value.

Here, R ₁ , R ₂ , R ₃ and R ₄ are bit values of the sub block in R data, G ₁ , G ₂ , G ₃ and G ₄ are bit values of the sub block in G data, B ₁ , B ₂ , B ₃ and B ₄ are bit values of the sub block in B data, m is the number of the vector values in each of the sub blocks, R _H is the upper representative value of the second vector representative value in R, R _L is a lower representative value of the second vector representative value. By receiving encoded data including a bitmap, vector representative values consisting of upper bits only, average distance between vector representative values and surrounding vector values, and correction positions of the respective vector values, And a decoder which decodes the encoded data.

Where (Rd, Gd, Bd) is the decoded pixel value, (Rr, Gr, Br) is the vector representative value, 'L' is the average distance, and (Rc, Gc, Bc) is the correction position, respectively. Indicates. Performing a VQ-BTC algorithm on the input data to generate first encoded data comprising a bitmap and a first vector representative value and adding flag bits for identifying the first encoded data to the first encoded data; A first encoder for adding one encoded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. A second encoder that generates the second vector representative values by using the average distance and the correction position, and adds a flag bit for identifying the second encoded data to the second encoded data; A selector for comparing the deviations between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result; And Decodes the selected encoded data to generate previous frame data, modulates the input data by selecting a modulation value listed in a lookup table using the previous frame data and the current frame data as addresses, and displays the modulated data on the LCD panel. And a driving circuit for the liquid crystal display panel. Performing a VQ-BTC algorithm on the input data to generate first encoded data comprising a bitmap and a first vector representative value and adding flag bits for identifying the first encoded data to the first encoded data; Adding 1 encoded data; Second coded data including the bitmap, second vector representative values consisting of upper bits only, an average distance between the second vector representative values and vector values in the vicinity thereof, and a correction position of each of the vector values. Generating; Correcting the second vector representative values using the average distance and the correction position; Adding a flag bit for identifying the second encoded data to the second encoded data; Comparing the deviations between the vector representatives with a predetermined threshold and selecting outputs of the first and second encoders according to the comparison result; And Decodes the selected encoded data to generate previous frame data, modulates the input data by selecting a modulation value listed in a lookup table using the previous frame data and the current frame data as addresses, and displays the modulated data on the LCD panel. A method of driving a liquid crystal display device, characterized in that.

Images