KR100530566B1 - Image compression coding and decoding apparatus using adaptive transformation method and method thereof - Google Patents

Abstract

The present invention provides an image compression encoding and decoding apparatus and method thereof used in a communication field or a multimedia field, and in particular, in case of outputting different transformation coefficients according to the conversion direction order of an input image signal block in image compression encoding and decoding. An image compression encoding and decoding apparatus using an adaptive transformation method that selects, encodes, and decodes a transform direction sequence according to a characteristic of the present invention, and a method thereof, comprising: an image compression encoding and decoding apparatus using an adaptive transformation method according to the present invention; According to the method, whether the signals in the block have high correlation in the vertical direction or high correlation in the horizontal direction by using the information of the coded block or the information of the coded block and the information about the block to be currently encoded. Determine whether you have a degree After orthogonal transform to the correlation direction is a great effect that can increase the efficiency of compression encoding By performing the orthogonal transform to a lower direction, an inverse orthogonal transformation by the orthogonal transformation in the opposite direction.

Description

The present invention relates to an image compression encoding and decoding apparatus and method for use in a communication field or a multimedia field. In particular, in the case of image compression encoding and decoding, a different conversion coefficient is output according to the conversion direction order of an input image signal block. The present invention relates to an image compression encoding and decoding apparatus using an adaptive transformation method of selecting, encoding, and decoding transformation direction orders according to characteristics of an image signal block, and a method thereof.

In general, in a method of compressing and encoding a digital image, a method of encoding an image signal block into a transform coefficient by using an orthogonal transform method and then encoding the transform coefficient is the most commonly used method. It is widely used in international standards for all digital compression encoding such as ISO / IEC JPEG, moving picture compression code and standardization method ISO / IEC MPEG, H.261 and H.263 of ITU-T.

Conventional converter image signal compression technology converts an input image signal block by an orthogonal transform method, quantizes a transform coefficient, transmits the quantized transform coefficient by variable length coding, and restores the signal through the reverse process. In the conventional video compression encoding and decoding system, as shown in FIG. 1, an encoding apparatus for encoding an input image signal block by using a compression technique, and an inverse encoding of the encoded signal by receiving the encoded signal. It consists of a decoding apparatus for restoring a signal through the process.

The encoding apparatus includes a subtractor 1 for outputting a prediction error signal block by subtracting an input image signal block and a reconstructed image signal block, an additional information signal (eg, shape information) and a variable length coded shape to assist in encoding. The additional information encoder 9 which outputs an information signal and the subtractor 1 receive the prediction error signal block and the shape information is selectively received by the switching operation of the switch 10a in the additional information encoder 9. An orthogonal transformer 2 for outputting a transform coefficient converted by a specific transform method, a quantizer 3 for receiving a transform coefficient from the orthogonal transformer 2 and converting the transform coefficient into a quantized transform coefficient block, and the quantizer ( A variable length coder 8 that receives the transform coefficient block quantized in 3) and converts the transform coefficient signal into a variable length coded transform coefficient signal, and the variable length coded signal from the variable length coder 8 Consists of a sub information encoder 9, the switch (10a) the shape information by switching operation of the input to receive selectively multiplexing the two signals a multiplexer 11 which is sent to the transmission medium 12 at.

On the other hand, the first inverse quantizer (4) for receiving and inverse quantization of the transform coefficient block quantized by the quantizer (3), and the inverse quantized transform coefficient block in the first inverse quantizer (4) An inverse orthogonal transformer 5 for receiving shape information selectively by the switching operation of the switch 10b in the additional information encoder 9 and converting the shape information into an inverse orthogonal transformed signal block; and an inverse orthogonal transformer 5 An adder 6a which adds an orthogonal transformed signal block and a reconstructed video signal block stored in the first memory 7 to be described later, and restores the video signal block, and receives a signal added by the adder 6a A first memory 7 is further constructed for storing and outputting the reconstructed video signal block to the subtractor 1.

The decoding apparatus receives a variable length coded signal from the multiplexer 11 through a transmission medium 12 and outputs a variable length coded shape information signal and a variable length coded input transform coefficient signal. (13), a decoder 15 for converting a variable length coded transform coefficient signal from the demux unit 13 into an input image block and converting the transform coefficient signal into a quantized transform coefficient block, and a quantized transform in the decoder 15 A second inverse quantizer 16 for receiving a coefficient block and converting the transform block into an inverse quantized transform coefficient block, an additional information decoder 14 for restoring variable length coded shape information in the demux unit 13, Inverse orthogonal transformer 17 receiving a dequantized transform coefficient block from two inverse quantizer 16 and selectively receiving shape information by a switching operation of switch 19 and outputting an inverse orthogonal transformed signal block At the same time, the inverse orthogonal transformer 17 receives the inverse orthogonal transformed signal block and at the same time receives the previous reconstructed image signal block stored in the second memory 18 to add the two signals to reconstruct the image signal block. An adder 6b for outputting, and a second memory 18 for storing the reconstructed video signal block outputted from the adder 6b and outputting the reconstructed video signal block to the adder 6b if necessary.

The shape information is information that divides an image into an object region and a non-object region (background). The shape information enables signal processing centering on an object region instead of the entire image in an encoder and a decoder. It takes the form of a binary mask that allows different values for pixels (backgrounds) rather than objects.

The operation of the conventional video compression encoding and decoding apparatus having the above configuration will be described.

First, when the subtractor 1 outputs the prediction error signal block obtained by subtracting the video signal block and the reconstructed video signal block, the quadrature converter 2 receives the prediction error signal block and at the same time the additional information encoder 9 receives the prediction error signal block. The shape information is selectively input, and the transform coefficient converted by the orthogonal transform method is output. At this time, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory 7 as well as the prediction error between the signal predicted from the first memory 7 and the current input signal. To this end, the value stored in the first memory 7 may be initialized to 0 to make the signal from the subtractor 1 equal to the input image signal.

In this case, the quantizer 3 receives the transform coefficient from the quadrature converter 2 and transforms the transform coefficient into a quantized transform coefficient block. The variable length encoder 8 receives the quantized transform coefficient block and converts the variable length coded transform into a quantized transform coefficient block. The signal is converted into a coefficient signal, and the variable length coded signal is output to the multiplexer 11.

The multiplexer 11 selectively receives the variable length coded transform coefficient signal and the variable length coded shape information from the side information encoder and multiplexes the two signals to output them through the transmission medium 12.

Subsequently, the demux 13 receives the encoded signal from the transmission medium 12 and outputs variable length coded shape information and a variable length coded transform coefficient block.

The decoder 15 receives the variable length coded signal from the demux unit 13 and converts it into a quantized transform coefficient block, and the second dequantizer 16 receives the quantized transform coefficient block and dequantizes it. Output the transform coefficient block.

The inverse orthogonal transformer 17 receives the dequantized transform coefficient block from the second inverse quantizer 16 and selectively receives shape information from the side information decoder 14, and outputs an inverse orthogonal transformed signal block. The adder 6b receives the inversely orthogonally transformed signal block and simultaneously receives the previously restored video signal block from the second memory 18 and adds the two signals to output the restored video signal block.

On the other hand, the transform coefficient block quantized by the quantizer 3 is input to the first inverse quantizer 4 and converted into a dequantized transform coefficient block, and the signal is input to the inverse orthogonal transformer 5.

At the same time, shape information is selectively inputted from the side information encoder 9 to the inverse orthogonal transformer 5, whereby the inverse orthogonal transformer 5 outputs an inverse orthogonal transformed signal block, and the adder 6a The inverse orthogonal transformed signal block and the image signal block previously restored in the first memory 7 are added to be converted into a restored image signal block, and the image signal block is stored in the first memory 7 and the It is input to the subtractor 1.

The additional information decoder 14 outputs the reconstructed shape information, and the second image 18 stores the reconstructed input image signal block for prediction.

On the other hand, the conversion method used in the quadrature converter 2 will be described with reference to Figs. 2A to 2E. Only the hatched portion is converted to the portion to be encoded and encoded. First, when the video signal block is input (FIG. 2A), the input signal is aligned with respect to the upper end of the unit block (FIG. 2B), and DCT transformed in the vertical direction (FIG. 2C).

Subsequently, when alignment is completed with respect to the left side, DCT conversion is performed in the horizontal direction (Fig. 2D) to obtain a final converted signal (Fig. 2E).

3A to 3C are diagrams for explaining that the transform coefficients are different according to the direction orthogonal transformed by the orthogonal transformer 2 of FIG. 1, and FIG. 3A is a signal block input to the orthogonal transformer 2, wherein the number The part without indicates the part without encoding, i.e., the part which does not perform encoding into the background area, and the part with the number is converted and encoded.

3B is a conversion coefficient when the first orthogonal transformation is performed in the longitudinal direction and then the orthogonal transformation is performed in the horizontal direction, and FIG.

Here, comparing FIG. 3B and FIG. 3C, it can be seen that different results, that is, different conversion coefficients are generated according to the orthogonal transformation direction order.

Looking at the details, the input signal block having a high similarity in the horizontal direction as shown in FIG. 3A is orthogonally transformed in the horizontal direction first and then orthogonally transformed in the vertical direction. As a result of FIG. 3B orthogonal transformation, the energy concentration is higher than that of FIG. 3B. Since only a small number of transformation coefficients are encoded, a higher compression performance is obtained.

That is, when orthogonal transformation is performed first in the vertical direction and then transversely orthogonally transformed as shown in FIG. 3B, 12 different values of coefficients of change must be encoded. Encoding efficiency can be increased because only coefficients having different values are encoded.

Conventionally, as shown in FIGS. 3A to 3C, without considering the similarity of an input signal that is orthogonally transformed, a method of unilaterally transforming after transverse orthogonal transformation or using a method of transverse orthogonal transformation after vertical orthogonal transformation There was a problem in that an optimum conversion result could not be obtained for the characteristics of the input signal.

Accordingly, the present invention has been made to solve the above problems, and an object of the present invention is to provide an apparatus and method for image compression encoding and decoding using an adaptive conversion method for increasing compression encoding efficiency.

To this end, an image compression encoding apparatus using an adaptive transformation method according to the present invention is an encoding apparatus that orthogonally transforms an input image signal block and compresses the same, and selectively inputs a transform coefficient obtained by orthogonal transformation and a current image block to be encoded. A conversion direction control signal for judging the correlation in the horizontal and vertical directions with respect to the input values and orthogonally converting the input image signal block in a highly correlated direction first and then orthogonally in a different direction from the input image signal block. The conversion mode control means for outputting and the orthogonal conversion means for receiving the conversion direction control signal from the conversion mode control means and performs orthogonal conversion correspondingly.

Another video compression encoding apparatus using an adaptive conversion method according to the present invention is an encoding apparatus that orthogonally transforms an input video signal block and compresses the received video signal block. Selectively receive the shape information of the two image signal blocks, and determine the correlation in the horizontal and vertical directions for the input values, and then orthogonally transform it in a direction having many correlations, and then A conversion mode control means for outputting a conversion direction control signal for performing orthogonal conversion in a direction; and an orthogonal conversion means for receiving a conversion direction control signal from the conversion mode control means and performing orthogonal conversion accordingly. . In this case, the conversion mode is controlled by receiving the shape information from the additional information encoder for the reconstructed video signal blocks and the shape information from the additional information encoder for the video signal block to be selectively encoded. Used as input to the means.

In addition, the video compression decoding apparatus using the adaptive conversion method according to the present invention is a decoding apparatus for decoding a signal transmitted by the compression coded video signal block, the decoding apparatus for receiving and decoding the transmitted compressed coded video signal In this case, a signal block that is currently transmitted and has undergone a variable length decoding process is selectively inputted, and a variable length decoded signal block is received previously and the correlation between the signal block is determined in the horizontal and vertical directions to determine the correlation. Inverse transform mode control means for outputting an inverse transform direction control signal for first inverse orthogonal transform and then inverse orthogonal transform in a different direction from the inverse transform mode; Characterized in that it consists of inverse orthogonal transformation means to perform.

An image compression encoding and decoding apparatus using an adaptive conversion method according to another embodiment of the present invention includes a subtractor for outputting a prediction error signal block by subtracting an input image signal block and a reconstructed image signal block, A shape information signal from the side information encoder for outputting shape information and a variable length coded shape information signal for assistance, and the additional information encoder for using only the prediction error signal block to be encoded in the subtractor and the shape information. Mode control means for selectively receiving a signal and outputting a corresponding conversion direction control signal, and receiving the prediction error signal block from the subtractor and selectively receiving variable length coded shape information from the side information encoder. The conversion direction is controlled to control the prediction error signal block. Orthogonal transform means for outputting a transform coefficient by performing an inter-transformation, a quantizer for receiving a quantized transform coefficient from the orthogonal transform means and outputting a quantized transform coefficient corresponding thereto, and receiving a variable quantized transform coefficient from the quantizer A variable length encoder for converting a length coded signal, a control signal encoder for receiving and encoding a conversion direction control signal from the mode control unit, a variable length coded signal from the variable length encoder, and receiving An encoder comprising a multiplexer which selectively receives shape information, receives a conversion direction control signal encoded by the control signal encoder, and multiplexes the three signals and transmits the three signals to a transmission medium; A demux unit for demultiplexing a signal input through a transmission medium in the encoding apparatus and outputting variable length coded shape information, a variable length coded quantized transform coefficient, and an encoded transform direction control signal; An additional information decoder for selectively receiving and decoding coded shape information, a decoder for variable length decoding receiving a variable length coded quantized transform coefficient from the demux unit, and a transform direction control signal encoded by the demux unit A control signal decoder to receive and decode a signal, a second inverse quantizer to dequantize the variable length decoded block by the decoder, and a dequantized transform coefficient block to receive the dequantized signal from the second inverse quantizer. Receiving shape information from the additional information decoder and converting the shape information from the control signal decoder A second inverse orthogonal transformation means for receiving a direction control signal and inversely orthogonally transforming the inverse quantized transform coefficient block in a direction opposite to an orthogonal transformation direction of the orthogonal transformation means, and inverse orthogonal transformation in the second inverse orthogonal transformation means An adder for receiving a previously restored video signal block and adding the two signals to output the restored video signal block, and storing the restored video signal block outputted from the adder and then storing the restored video signal block. Characterized in that it consists of a decoding device consisting of a memory to output for prediction.

In addition, the video compression encoding and decoding apparatus using the adaptive conversion method according to another embodiment of the present invention includes a subtractor for outputting a prediction error signal block by subtracting the input image signal block and the reconstructed image signal block, Shape information is input from the side information encoder to use only the prediction error signal block to be encoded by the subtractor and the shape information signal for outputting shape information and variable length coded shape information signal. A mode control means for receiving a corresponding control direction control signal and a prediction error signal block received from the subtractor, and optionally receiving shape information from the additional information encoder to control the conversion direction by the mode control means. Outputs the conversion coefficient by orthogonally transforming the prediction error signal block A transform unit, a quantizer that receives a transform coefficient from the orthogonal transform unit and quantizes the transform coefficient, and outputs a quantized transform coefficient corresponding to the transform coefficient; A variable length encoder for converting, a control signal encoder for receiving and encoding a conversion direction control signal from the mode control means, and the mode control means and the control signal encoder according to whether to use the conversion direction control signal during orthogonal or inverse orthogonal conversion. A mode controller control signal generating means for generating a mode controller control signal for controlling the on / off mode of the controller; a mode controller control signal encoder for receiving and encoding a mode controller control signal from the mode controller control signal generating means; Receiving a variable length coded signal from the length encoder, The shape information is selectively received from the side information encoder, the conversion direction control signal encoded by the control signal encoder is received, the mode controller control signal encoded by the mode controller control signal encoder is received, multiplexed, and transmitted to the transmission medium. An encoding device comprising a multiplexer; A demux unit for demultiplexing a signal input through a transmission medium in the encoding apparatus and outputting variable length coded shape information, a transform coefficient, a conversion direction control signal, and a mode controller control signal, and a shape encoded by the demux unit An additional information decoder to selectively receive and decode information, a variable length decoder to receive a variable length coded video signal block in the demux unit, and a variable length coded conversion direction control signal in the demux unit; A control signal decoder to receive and decode, a mode controller control signal decoder to receive and decode a mode controller control signal encoded by the demux unit, and to control an on / off mode of the control signal decoder, and to perform a variable length decoding on the decoder. A second inverse quantizer for receiving and inversely quantizing a signal block, and the second inverse quantizer An inverse quantized transform coefficient block, a shape information received from the additional information decoder, a conversion direction control signal from the control signal decoder, and the orthogonal transform means A second inverse orthogonal transform means for inverse orthogonal transform and a signal block that is inversely orthogonal transformed in the second inverse orthogonal transform means, and a previously reconstructed image signal block to receive the two signals; And a decoder configured to add and output the reconstructed image signal block, and a memory configured to store the reconstructed image signal block output from the adder and output the prediction signal to the adder for prediction.

On the other hand, the video compression encoding method using the adaptive conversion method according to the present invention is a coding method for orthogonally transforming an input video signal block, the transform mode control means is to perform the current encoding and the transform coefficient obtained by orthogonal transformation previously A conversion direction control signal for selectively receiving an image signal block and determining the correlation in the horizontal and vertical directions with respect to the input values, and performing orthogonal transformation in a direction having a high correlation, and then performing orthogonal transformation in a different direction. A conversion direction control signal generating step of outputting a; and orthogonal conversion means receives the conversion direction control signal, characterized in that it consists of an orthogonal conversion step of performing orthogonal conversion sequentially accordingly.

In addition, the video compression decoding method using the adaptive conversion method according to the present invention is a method of receiving and decoding a compressed coded video signal transmitted, wherein the inverse transform mode control means selectively receives a signal block that is currently variable-length decoded. Receive the variable-length decoded signal block and determine the correlation in the horizontal and vertical directions for the signal blocks, and then inverse orthogonal transform in the direction with less correlation and inverse orthogonal transform in the other direction. Generating an inverse transform direction control signal for outputting an inverse transform direction control signal, and an inverse orthogonal transform step in which an inverse orthogonal transform means receives the inverse transform direction control signal and sequentially performs inverse orthogonal transform corresponding thereto; .

In addition, the video compression encoding and decoding method using an adaptive conversion method according to another embodiment of the present invention is a method of encoding and decoding an input video signal block, using orthogonal transformation using an image signal block to be encoded and shape information. And a conversion direction control signal generation step of generating a conversion direction control signal for determining a conversion direction order of the inverse orthogonal conversion, and a conversion direction control signal generated in the conversion direction control signal generation step, while converting the conversion direction control signal. An encoding step of orthogonally transforming an image signal block to be encoded by quantization and variable length encoding, a transmission step of multiplexing and transmitting the transform direction control signal and the variable length encoded image signal block encoded in the encoding step, and Receives and demultiplexes the encoded signal transmitted in the transmission step. A signal separation step of separating the encoded conversion direction control signal and the variable length coded video signal block, and decoding the encoded conversion direction control signal separated in the signal separation step, while variable length encoding the variable length coded video signal block A decoding step of decoding and inverse quantization, and an inverse orthogonal transform direction order is determined by the transform direction control signal decoded in the decoding step to inverse orthogonal transform the inverse quantized transform coefficient block and convert the previously reconstructed image signal block It is characterized by consisting of a signal recovery step for restoring the video signal block initially input.

Hereinafter, an image compression encoding and decoding apparatus using an adaptive conversion method according to an embodiment of the present invention, and a method thereof will be described in detail with reference to the accompanying drawings.

4 is a control block diagram of an image compression encoding and decoding apparatus using an adaptive transform method according to an embodiment of the present invention.

A coding apparatus according to the present invention will be described with reference to FIG. 4.

The subtractor 1 subtracts the image signal block and the reconstructed image signal block to output the prediction error signal block, and the side information encoder 9 outputs the shape information to help the encoding. At this time, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory 7 as well as the prediction error between the signal predicted from the first memory 7 and the current input signal. To this end, the value stored in the first memory 7 may be initialized to 0 to make the signal from the subtractor 1 equal to the input image signal.

In addition, the orthogonal transformation means 1000 receives a prediction error signal block from the subtractor 1, and selectively receives shape information from the additional information encoder 9, orthogonally by the conversion control means 100 to be described later. The conversion mode is controlled to perform orthogonal conversion and output the conversion coefficient.

According to an exemplary embodiment of the orthogonal transformation means 1000, as shown in FIG. 5A, a prediction error signal block is selectively input by the subtractor 1, and shape information is selectively input by the additional information encoder 9. A DCT transformation of the prediction error signal block in the vertical direction and DCT conversion in the horizontal direction, and a prediction error signal block selectively received from the subtractor 1, and the additional information encoder An orthogonal transformer 2 (1140) which selectively receives shape information at (9) and performs DCT conversion on the prediction error signal block in the horizontal direction and then DCT conversion in the vertical direction. In this case, the prediction error signal block may be initialized to 0 to be identical to the input image signal block.

The orthogonal transform means 1000 receives a conversion direction control signal from the conversion mode control means 100 and performs a switching operation so that the predictive error signal block output from the subtractor 1 is the orthogonal transformer 1 1120 or the like. The first switch 1110 to control input to the quadrature converter 2 (1140) and the conversion direction control signal from the conversion mode control means 100 are switched to operate the quadrature converter 1 (1120) or the quadrature converter. A second switch 1130 is further configured to intercept the conversion coefficient output from 2 1140 to the first quantizer 3.

On the other hand, in another embodiment of the orthogonal conversion means 1000, the conversion mode control means 100 receives the conversion direction control signal from the conversion mode control means 100 as shown in FIG. While receiving the input from the preprocessor 1210 and the preprocessor 1210 which rotates the prediction error signal block input by the subtractor 1 by 90 degrees counterclockwise and outputs the prediction error signal block as it is in the off mode. An orthogonal converter 1220 for selectively receiving shape information from the additional information encoder 9 and performing DCT conversion in a vertical / horizontal direction and a DCT conversion in a horizontal / vertical direction; and a conversion direction in the conversion mode control means 100 When the on / off mode is controlled by receiving the control signal block, the quadrature converter 1220 receives the conversion coefficient and rotates 180 degrees around the axis connecting the upper left and lower right sides of the unit block. And a post processor 1230 for outputting the transform coefficient block as it is. In this case, the prediction error signal block may be initialized to 0 to be equal to the input image signal block.

In addition, the first quantizer 3 serves to receive the transform coefficient from the orthogonal transform means 1000 and convert the transform coefficient into a quantized transform coefficient block, and the variable length encoder 8 is the first quantizer 3. Converts the quantized transform coefficient block into a variable length coded signal, and the multiplexer 11 receives the variable length coded signal from the variable length coder 8 while the side information encoder 9 ) Selectively receives the variable-length coded shape information signal and multiplexes the two signals to transmit them to the transmission medium.

In one embodiment of the conversion mode control means 100, the first quantizer 3 receives a quantized transform coefficient, and the subtractor 1 selectively receives a prediction error signal corresponding thereto. Outputs the conversion mode control signal. In this case, the prediction error signal may be made equal to the input image signal block by initializing the value stored in the first memory to zero.

In addition, the detailed configuration of the conversion mode control means 100 is a memory 110 for receiving and storing the quantized transform coefficients in the first quantizer 3, and inputs the quantized transform coefficients in the memory 110 On the other hand, it consists of a mode controller 120 for selectively receiving the prediction error signal block from the subtractor 1 and outputs a corresponding conversion direction control signal. In this case, the prediction error signal block may be initialized to 0 to be identical to the input image signal block.

On the other hand, the conversion mode control means 100 is another embodiment, in which the memory 110 receives a transform coefficient quantized by the first quantizer 3 in place of the first memory 7 which will be described later. The reconstructed video signal block and the prediction error signal block to be encoded by the subtractor 1 are selectively inputted, and the reconstructed video signal block by the additional information encoder 9 to use only the prediction error signal in the shape information. The shape information of the prediction error signal block to be encoded may be selectively input by the switching action of the switch 10g, and output a conversion direction control signal accordingly. In this case, the prediction error signal block may be initialized to 0 to be identical to the input image signal block.

In addition, the first inverse quantizer 4 receives the transform coefficient block quantized by the first quantizer 3 and converts the inverse quantized transform coefficient block into a first inverse orthogonal transform means 140. Receives an inverse quantized transform coefficient block from the first inverse quantizer 3, selectively receives shape information from the additional information encoder 9, and the inverse transform mode is changed by the conversion mode control means 100. It is controlled to output the corresponding inverse transform coefficient.

As shown in FIG. 5B, the first inverse orthogonal transform unit 140 selectively receives the dequantized transform coefficient block from the first inverse quantizer 4, and the additional information encoder ( 9) an inverse orthogonal transformer 1 (2120) and a first inverse quantizer (4) for selectively receiving shape information and inversely transforming the inverse quantized transform coefficient block in the vertical direction, and then inversely transforming the DCT in the horizontal direction. Inverted orthogonal transformer 2 (2140) which selectively receives the inverse quantized transform coefficient block at the same time, inversely transforms the DCT in the horizontal direction after selectively receiving shape information from the side information encoder 9 Consists of

In addition, the first inverse orthogonal transformation means 140 receives a conversion direction control signal from the conversion mode control means 100 so that a switching operation is controlled so that the inverse quantized transformation is output from the first inverse quantizer 4. A third switch 2110 for intermitting the coefficient block input to the inverse orthogonal transformer 1 (2120) or the inverse orthogonal transformer 2 (2140) and a switching direction control signal from the conversion mode control means (100) for switching The fourth switch 2130 is further configured to control an operation so that an inverse transform coefficient output from the inverse orthogonal transformer 1 (2120) or the inverse orthogonal transformer (2140) is input to the first adder (6a). It is.

On the other hand, the first inverse orthogonal conversion means 140 is a mode in which an on / off mode is controlled by receiving a conversion direction control signal from the conversion mode control means 100 as shown in FIG. 6B as another embodiment. The first inverse quantizer 4 receives the inverse quantized transform coefficient block and rotates 180 degrees around the axis connecting the upper left and lower right corners of the unit block and outputs the dequantized transform coefficient signal block as it is in the off mode. Preprocessor 1 (2210) and the preprocessor 1 (2210), while the additional information encoder (9) selectively receives the shape information inverted DCT conversion in the vertical / horizontal direction, and then the horizontal / vertical direction Inverted orthogonal converter 3 (2202) for inverse DCT conversion, and the inverse orthogonal converter 3 (2220) in the on mode when the on / off mode is controlled by receiving the conversion direction control signal from the conversion mode control means 100 Modulus Power take consists of one processor 2230 and then to a 90 degree rotation in the clockwise direction to output a reverse conversion factor and output as an off-mode when reverse conversion coefficient.

In addition, the first adder 6a receives the inverse transform coefficients from the first inverse orthogonal transformation means 140 and simultaneously receives the previously restored image signal blocks stored in the first memory 7 described later. Adds a signal to restore the image signal block, and the first memory 7 receives the reconstructed image signal block from the adder 6a, stores the image signal block for the next prediction, and restores the image signal to the adder 1 Outputs the signal block.

In addition, a switch 10b is interposed between the subtractor 1 and the mode controller 120 to control the prediction error signal block output from the subtractor 1 from being input to the mode controller 120. Between the additional information encoder 9, the orthogonal transform means 1000, and the first inverse orthogonal transform means 140, the orthogonal transform means 1000 and the first inverse orthogonal transform means 140 in the additional information encoder 9 The switch 10c which controls an input of shape information is comprised.

In addition, between the additional information encoder 9 and the multiplexer 11, a switch 10a for intermitting the shape information output from the additional information encoder 9 to the multiplexer 11 is configured. have.

On the other hand, the decoding apparatus according to the present invention will be described with reference to FIG.

The demux unit 13 receives the coded signal from the encoding device and demultiplexes the coded shape information and outputs the encoded shape information and the encoded image signal. The decoder 15 is connected to the demux unit 13. It receives the encoded video signal and converts it into a variable length decoded signal.

The second inverse quantizer 200 receives a variable length decoded signal from the decoder 13 and converts the dequantized transform coefficient block into an inverse quantized transform coefficient block. 13) and selectively receives the shape information encoded in the function to restore shape information.

In addition, the second inverse orthogonal transformation means 2000 receives the dequantized transform coefficient block from the second inverse quantizer 200 and selectively receives shape information restored from the side information decoder 14. The inverse transform mode is controlled by the inverse transform mode control means 210, which will be described later, and performs inverse orthogonal transformation, thereby outputting an inverse transform coefficient.

According to an embodiment of the second inverse orthogonal transformation means 2000, as shown in FIG. 5B, the second inverse quantizer 200 selectively receives the dequantized transform coefficient block, and the additional information decoder ( An inverse orthogonal transformer 1 '(2120) for performing inverse DCT conversion on the inverse quantized transform coefficient block in a vertical direction and selectively inverse DCT in the horizontal direction by selectively receiving the shape information restored in (14); A quantizer 200 selectively receives an inverse quantized transform coefficient block, while selectively receiving shape information reconstructed by the side information decoder 14 to inversely transform the inverse quantized transform coefficient block in a horizontal direction. Then, it is composed of an inverse orthogonal transformer 2 '(2140) for inverse DCT conversion in the longitudinal direction.

In addition, the second inverse orthogonal transformation means 2000 receives the inverse transformation direction control signal from the inverse transformation mode control means 210 and controls a switching operation to convert the quantized inverse quantized block in the second inverse quantizer 200. A third 'switch 2110 for controlling input to the inverse orthogonal transformer 1' (2120) or an inverse orthogonal transformer 2 '(2140), and an inverse conversion direction control signal from the inverse conversion mode control means 210; Switching operation is controlled to control the inverse transform coefficient output from the inverse orthogonal transformer 1 '(2120) or the inverse orthogonal transformer 2' (2140) 2140 (2b) 4 'switch (2130) Is further provided.

On the other hand, according to another embodiment of the second inverse orthogonal transformation means 2000, the inverse transformation direction control signal is input from the inverse transformation mode control means 210 as shown in FIG. The inverse quantized transform coefficient block input from the second inverse quantizer 200 is rotated 180 degrees around the axis connecting the upper left and the lower right of the unit block and outputs the inverse quantized signal block input in the off mode as it is. The preprocessor 1 '2210 and the preprocessor 1' 2210 are input, while the additional information decoder 14 selectively receives the shape information signal currently restored, and performs reverse DCT conversion in the vertical and horizontal directions. The inverse orthogonal converter 3 '(2220) for inverse DCT conversion in the horizontal / vertical direction, and the inverse orthogonal converter when the on / off mode is controlled by receiving an inverse conversion direction control signal from the inverse conversion mode control means 210; 3 '(2220) ) Is a post-processor 1 '(2230) is rotated 90 degrees in the clockwise direction to receive the inverse transform coefficient in the output mode in the off mode.

In addition, the second adder 6b receives a signal block inversely transformed by the second inverse orthogonal transformation means 2000 and simultaneously receives a reconstructed image signal block to be described later, and adds the two signals to reconstruct the image signal block. The second memory 18 stores the reconstructed input image signal block for prediction, and the signal block is output to the second adder 6b for the next prediction.

In addition, the inverse transform mode control unit 210 receives a variable length decoded signal from the decoder 15 and outputs a corresponding inverse transform direction control signal. The configuration of the variable length decoding in the decoder 15 is performed. A memory 211 for receiving and storing the received signal and a variable length decoded signal previously received from the memory 211, and selectively receiving a current variable length decoded signal from the decoder 15 accordingly. And a mode controller 213 for outputting an inverse conversion mode control signal.

On the other hand, the inverse conversion mode control means 210 is another embodiment, the image signal block restored in the second memory 18 to be described later instead of the configuration in which the memory 211 receives the variable-length decoded signal previously In order to use only the prediction error signal in the shape information and the shape information restored by the additional information decoder, the decoder 15 selectively receives the current variable length decoded signal and receives the corresponding inverse transform mode control signal. Can be configured to output.

On the other hand, between the additional information decoder 14 and the second inverse orthogonal transformation means 2000, the shape information output from the additional information decoder 14 is intermittently inputted to the second inverse orthogonal transformation means 2000. The switch 16 is further comprised.

Further, between the decoder 15 and the mode controller 213, a switch 17 for intermittently inputting a variable length decoded signal from the decoder 15 to the mode controller 213 is further configured. Between the demux 13 and the additional information decoder 14, a switch 10e intermittently intercepting the coded shape information output from the demux 13 to the additional information decoder 14. It is further provided.

An encoding method and a decoding method using an image compression encoding apparatus and a decoding apparatus using the adaptive transformation method having the above-described configuration will be described.

First, the coding method will be described.

The image signal block is input to the subtractor 1, and the shape information is input to the additional information encoder 9 and encoded.

At this time, the subtractor 1 subtracts the input image signal block and the reconstructed image signal block output from the first memory 7 to obtain a prediction error signal block, and outputs the signal to the orthogonal transformation means 1000.

In this case, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory 7 as well as the prediction error between the signal predicted from the first memory 7 and the current input signal. To this end, the value stored in the first memory 7 may be initialized to 0 to make the same as the signal from the subtractor 1 and the input image signal.

The orthogonal transformation means 1000 receives a prediction error signal block from the subtractor 1, and selectively receives shape information from the side information encoder 9, and the conversion mode is controlled by the conversion mode control means 100. Outputs the conversion coefficient by performing orthogonal transformation.

Thereafter, the first quantizer 3 converts the transform coefficient output from the orthogonal transform means 1000 into a quantized transform coefficient block, and the variable length coder 8 is quantized by the first quantizer 3. A transform coefficient block is received and converted into a variable length coded transform coefficient signal, and the multiplexer 11 receives the variable length encoded transform coefficient signal and at the same time the shape information signal variable length coded by the additional information encoder 9. Selectively receives and transmits the two signals through the transmission medium 12. FIG.

Meanwhile, the quantized transform coefficient block output from the first quantizer 3 is converted into an inverse quantized transform coefficient block which is input to the first inverse quantizer 4 and the first inverse orthogonal transform means 140 The first inverse quantizer 4 receives an inverse quantized transform coefficient block, and selectively receives shape information encoded by the side information encoder 9 by an inverse transform mode by the conversion mode control means 100. Is controlled to output inverse transformation coefficient by performing inverse orthogonal transformation.

Subsequently, the first adder 6a receives an inverse transform coefficient from the first inverse orthogonal transformation means 140 and receives an image signal block reconstructed from the first memory 7 and adds the two signals to add an image signal block. The image signal block is restored and stored in the first memory 7, and is restored to the subtractor 1.

In addition, the method for generating the conversion direction control signal in the mode controller 120 of the conversion mode control means 100 can be implemented in two ways.

The first method is a method of outputting a conversion direction control signal when the block unit coding method is used as shown in FIG.

Here, the block unit encoding method refers to a method of dividing an input signal into blocks having a predetermined size and then performing transform encoding on the basis of the divided blocks.

X is a block to be encoded currently input from the subtractor 1, and is not involved in generating a conversion direction control signal. A, B, and C are output from the first quantizer 3 and stored in the memory 110. A quantized transform coefficient input to the controller 120, wherein A is a block located on the left side of the block X, B is a block located on the upper left side in the block X, and C is a block located above the block X.

Using the rate of change of DC values among the transform coefficients of each block to determine the mode, and the rate of change in the horizontal and vertical directions to obtain the mode is expressed by Equation (1).

[Equation 1]

Here, Grad_Vert represents the rate of change in the vertical direction, and Grad_Hori represents the rate of change in the horizontal direction.

Here, ABS (a) represents the absolute value of a, and finally outputs a value for a direction in which the conversion direction control signal change rate input to the orthogonal transformation means 1000 is small.

That is, if Grad_Vert is less than or equal to Grad_Hori, orthogonal transformation is performed first in the vertical direction, and then the conversion direction control signal is output so as to be orthogonal in the horizontal direction. Outputs a conversion direction control signal to orthogonally convert in the direction.

The meaning of this method is as follows. The DC values of each block mean the average of each block, and the rate of change between the average values of the blocks to be encoded is the property of the current block, that is, whether the signals in the block have a high correlation in the horizontal direction. It shows information about whether it has a high correlation. Therefore, it is possible to efficiently determine the mode for the conversion direction only by using information on the neighboring blocks.

In addition, since this method uses only information on blocks that have already been encoded, there is an advantage that a transform direction control signal for a block to be currently encoded can be obtained by a simple operation.

In addition, since only the blocks that are already coded are used, the decoder has the advantage of determining the same control signal determined by the encoder without additionally transmitting a control signal to the decoder.

Another method is to use the information of transform coefficients already quantized around the block to be encoded, instead of the method of FIG. 7, in addition to the transform coefficient block already quantized so that the control mode can be determined more accurately as shown in FIG. The difference is that the information about the block to be encoded is used together.

In this case, the value of the quantized DC signal of the block to be currently encoded is used so that the control signal is not additionally transmitted to the decoder.

In FIG. 8, X is a block to be encoded currently input from the subtractor 1, and A, B, and C are output from the first quantizer 3, stored in the memory 110, and then input into the mode controller 120. Is a block located on the left in block X, B is a block located in the upper left corner in block X, and C is a block located above block X.

For the mode decision, the rate of change of the DC values, such as Equation 2, among the transform coefficients of each block is used.

[Equation 2]

Here, ABS (a) represents the absolute value of a, and in order to obtain the mode, the rate of change in the horizontal and vertical directions is calculated as in Equation (3).

[Equation 3]

Here Grad_Vert represents the rate of change in the vertical direction, Grad_Hori represents the rate of change in the horizontal direction.

Finally, the conversion direction control signal for the quadrature converter outputs a value for the direction in which the rate of change is small.

That is, if Grad_Vert is less than or equal to Grad_Hori, orthogonal transformation is performed first in the vertical direction, and then the conversion direction control signal is output so as to be orthogonal in the horizontal direction. Outputs a conversion direction control signal to orthogonally transform in the direction.

This method uses the information of the current block together with the information of the neighboring blocks, which has the advantage of outputting a more accurate conversion direction control signal than in Fig. 7, but in order to obtain the DC of the current block and the rate of change, more operations are needed than in Fig. 7. There is a necessary disadvantage. In addition to the addition operation, various methods for determining the characteristics of the signal, such as a maximum value and a subtraction operation, may be used to obtain the change rate.

In addition, since the orthogonal transformation means 1000 can be realized as two embodiments as described above, the conversion method performed by the orthogonal transformation means 1000 will be described individually for each embodiment.

Orthogonal transformation method made in the orthogonal transformation means (Fig. 5a) which is the first embodiment,

First, the conversion mode control means 100 inputs a conversion direction control signal for determining the conversion mode to the first and second switches 1110 and 1130. In this case, if the conversion direction control signal is vertical orthogonal, If the conversion information is included, the first and second switches 1110 and 1130 switch the input and output terminals to the quadrature converter 1 1120. Accordingly, the quadrature converter 1 1120 inputs the predictive error signal block from the subtractor 1. At the same time, the additional information encoder 9 selectively receives shape information, and outputs a transform coefficient signal obtained by orthogonal transformation of the inputted prediction error signal block in the vertical direction and then orthogonally transformed in the horizontal direction. In this case, the prediction error signal block may be initialized to 0 to be equal to the input image signal block.

On the other hand, if the input conversion direction control signal has the vertical orthogonal transformation information after the transverse orthogonal transformation, the first and second switches 1110 and 1130 switch the input and output terminals to the orthogonal transformer 2 (1140), accordingly, Orthogonal transformer 2 (1140) receives the prediction error signal block from subtractor 1 and selectively receives shape information from side information encoder (9). Outputs the orthogonal transform coefficient.

Orthogonal transformation method made in the orthogonal transformation means (Fig. 6a) of the second embodiment,

First, a conversion direction control signal for determining a conversion mode is input by the conversion mode control means 100, and if the mode information is a vertical orthogonal conversion mode after horizontal orthogonal conversion, the preprocessor 1210 is operated to subtract the subtractor (1). Rotates the signal input at 90 degrees counterclockwise, and the orthogonal converter 1220 receives the rotated signal and at the same time selectively receives the shape information from the additional information encoder (9) vertical orthogonal transformation and then transverse orthogonal transformation Input to postprocessor 1230.

At this time, the post-processor 1230 is operated to rotate the orthogonal signal 180 degrees around the axis connecting the upper left and lower right of the unit block to output a conversion coefficient.

On the other hand, if the mode is a transverse orthogonal transformation after the vertical orthogonal transformation, the signal is output as it is without operating the preprocessor 1210 and the post processor 1230, and only the orthogonal transformer 1220 is operated to be input from the subtractor 1. A transform coefficient obtained by transverse orthogonal transformation after vertical orthogonal transformation of the predicted error signal block is output. In this case, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory 7 as well as the prediction error between the signal predicted from the first memory and the current input signal. 1 The value stored in the memory can be initialized to 0 to make it the same as the signal from the subtractor and the input video signal.

On the other hand, since the first inverse orthogonal transformation means 140 can be realized as two embodiments as described above, the inverse orthogonal transformation method will be described individually for each embodiment.

The inverse orthogonal transformation method made by the first inverse orthogonal transformation means 140, which is the first embodiment,

First, the conversion mode control means 100 inputs a conversion direction control signal to the third and fourth switches 2110 and 2130. If the conversion direction control signal has vertical orthogonal transformation information after horizontal orthogonal transformation, The third and fourth switches 2110 and 2130 switch the input and output terminals to the inverse orthogonal transformer 1 (2120). Accordingly, the inverse orthogonal transformer 1 (2120) is inversely quantized by the first inverse quantizer (4). At the same time as receiving the block, the side information encoder 9 selectively receives the shape information, and inversely orthogonally transforms the input inverse quantized transform coefficient block in the vertical direction, and inversely orthogonally in the horizontal direction to output the corresponding inverse transform coefficient. .

On the other hand, if the input conversion direction control signal has transverse orthogonal transformation information after vertical orthogonal transformation, the third and fourth switches 2110 and 2130 switch the input and output terminals to the inverse orthogonal transformer 2 (2140). Inverse orthogonal transformer 2 (2140) receives the inverse quantized transform coefficient block from the first inverse quantizer (4) and simultaneously receives the shape information from the side information encoder (9). After the inverse orthogonal transformation in the horizontal direction and the inverse orthogonal transformation in the vertical direction and outputs the corresponding inverse transform coefficient.

The inverse orthogonal transformation method made by the inverse orthogonal transformation means (FIG. 6B) of the second embodiment,

First, a conversion direction control signal for determining a conversion mode is input by the conversion mode control means 100, and if the mode information is the vertical orthogonal conversion mode after the transverse orthogonal conversion, the preprocessor 1 2210 is operated to operate the first reverse direction. The inverse quantized transform coefficient block input from the quantizer 4 is rotated 180 degrees about the axis connecting the upper left and lower right of the unit block, and the inverse orthogonal transformer 3 2220 receives the rotated signal and adds the same. The information encoder 9 selectively receives shape information, and then transverse inverse orthogonal transformation and vertical inverse orthogonal transformation to input the post-processor 1 (2230).

In this case, the post processor 1 2230 is operated to output an inverse transform coefficient obtained by rotating the inverse orthogonal transformed signal block by 90 degrees clockwise.

On the other hand, if the mode is a transverse orthogonal transformation after the vertical orthogonal transformation, the first inverse quantizer 4 is operated by operating only the inverse orthogonal transformer 2220 without operating the preprocessor 1 2210 and the post processor 2230. An inverse transform coefficient formed by transverse inverse orthogonal transformation of the input inverse quantized transform coefficient block and then inverse inverse orthogonal transformation is output.

Meanwhile, the decoding method will be described.

First, the demux 13 receives the decoded video signal through the transmission medium 12 to demultiplex and output variable length coded shape information and a variable length coded transform coefficient signal, and the decoder 15 The variable length coded image signal is received and converted into a quantized transform coefficient signal, while the additional information decoder 14 selectively receives the variable length coded shape information and restores the original shape information.

In this case, the second inverse quantizer 200 receives the transform coefficient signal quantized by the decoder 15 and converts the inverse quantized transform coefficient block, and the second inverse orthogonal transformation means 2000 performs the second inverse quantization. While receiving the inverse quantized transform coefficient block from the device 200, the shape information restored by the additional information decoder 14 is selectively received, and the inverse transform direction control signal is received from the inverse transform mode control means 210. Is controlled to output the inverse orthogonal transformed signal block.

Subsequently, the second adder 6b receives the signal block inversely orthogonally transformed by the second inverse orthogonal transformation means 2000 and the image signal block previously restored in the second memory 18, and adds the two signals. Output the reconstructed video signal block.

Meanwhile, the second memory 18 receives and stores the reconstructed video signal block output from the second adder 6b and outputs the reconstructed video signal block to the second adder 6b.

The method of generating the inverse conversion direction control signal by the inverse conversion mode control means 210 is the same as the conversion mode control means 100 described above, and the conversion mode control means 100 uses only the DC value to convert the conversion direction control signal. In the same manner, the inverse transform mode control means 210 generates the inverse transform direction control signal using the DC value, while using the image signal block and the shape information restored by the transform mode control means 100. In 210, the restored video signal block and shape information should be used.

Meanwhile, the method of inverse orthogonal transformation in the second inverse orthogonal transformation means 2000 is the same as the method of the first inverse orthogonal transformation means 140 described above.

Meanwhile, FIG. 9 is a diagram illustrating an input image having star shape information, in which a hatched portion becomes an object portion of the input image and the remaining portion is a background portion of the input image.

10 is a diagram illustrating a block for encoding an input image having the shape information of FIG. 9. Only the hatched block is encoded, and the remaining blocks are excluded from the encoding target.

In order to select a transform method for a block to be encoded and not transmit the information to the decoder, the transform method should be selected using only information on a block that is already encoded.

That is, the conversion method should be selected using only the blocks located on the left in the same row as the blocks located above the block to be encoded.

Therefore, if there is no already encoded block, such as a hatched block in FIG. 11 or insufficiently enough to select a transform method, the transform method for the block to be encoded cannot be selected. In this case, it is assumed that the encoding order starts from the upper left block and runs in the right direction to the lower right block.

Therefore, when the conversion method is directly selected using the signal to be encoded, the current signal to be encoded is directly analyzed and thus the accuracy of selecting the conversion method is higher than the method of determining the conversion method using only the surrounding signals. I can guarantee it. Therefore, in spite of generation of additional bits due to transmission of the mode control signal, there is an excellent effect of increasing high coding efficiency by directly using a signal to be currently encoded. In addition, when determining the conversion method using the peripheral signal, there is an advantage that can solve the problem of the bitstream editability (bitstream editability) described later.

12 is a control block diagram of an image compression encoding and decoding apparatus using an adaptive transformation method according to the second embodiment of the present invention, focusing on the above points. Description is omitted.

In the second embodiment, the conversion mode control means 100 for generating the conversion direction control signal used in the first embodiment, and the switch 10b intermittently inputting the prediction error signal block to the conversion mode control means 100. Instead of the switch 17 which controls the inverse transform mode control means 210 for outputting the inverse transform direction control signal and the variable length decoded signal from the decoder 15 to the inverse transform mode control means 210 as follows: The component is added.

The component added to the second embodiment selectively receives shape information input through the additional information encoder 9 and uses the predictive error signal input through the subtractor 1 to use only the prediction error signal in the shape information. A mode control means 3000 for generating a conversion direction control signal for receiving a block and determining an orthogonal conversion direction order using the block; and a control signal encoder 4000 for receiving and encoding a conversion direction control signal from the mode control means 3000. And variable length decoding of the encoded transform direction control signal output from the control signal encoder 4000 through the multiplexer 11, the transmission medium 12, and the demux unit 13. A control signal decoder 5000 for inputting the conversion direction control signal generated at 3000 to the second inverse orthogonal conversion means 200.

In this case, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory 7 as well as the prediction error between the signal predicted from the first memory 7 and the current input signal. To this end, the value stored in the first memory 7 may be initialized to 0 to make the same as the signal from the subtractor 1 and the input image signal.

In addition, the mode control means 3000 may be configured in three embodiments, which will be described with reference to FIGS. 13, 14, and 15, respectively.

FIG. 13 is a diagram illustrating Embodiment 1 of a mode control means. The mode control means according to Embodiment 1 receives a two-dimensional memory 3100 that receives and stores a prediction error signal block to be currently encoded, which is input from the subtractor 1. ), And a plurality of horizontal subtractors 3110 for repetitively obtaining the differences between pixel values existing in one row of the prediction error signal block stored in the two-dimensional memory 3100 and the horizontal subtractor 3110. A plurality of horizontal adders 3120 for calculating the sum of the obtained subtraction values, and a final evaluator for adding the sum of the subtraction values for the currently obtained one row output from the horizontal adder 3120 and the subtraction values of the previously obtained rows. (3150), a first temporary memory (3160) for temporarily storing the addition value for the row output from the final evaluator 3150, and the output of the last horizontal adder (3150) for the last row; 1 rain And a switch (3115-a) which is switched to operation as an input to the group (3190).

In addition, the mode control means according to the first embodiment includes a plurality of vertical subtractors 3130 for repeatedly calculating the difference between pixel values existing in one column of the prediction error signal block stored in the two-dimensional memory 3100 for each column; The sum of the plurality of vertical adders 3140 that calculate the sum of the obtained subtraction values in the vertical subtractor 3130, and the sum of the subtraction values for the currently obtained one row output from the vertical adder 3140 and the subtracted values of the previously obtained columns. A final vertical adder 3170 for adding the sum, a second temporary memory 3180 for temporarily storing an addition value for the column output from the final vertical adder 3170, and the final vertical adder 3140 for the last column A switch 3115-b for switching to make an output of the first comparator 3190 an input to the first comparator 3190, and an addition value for a row and a column for the rows output from the first and second temporary memories 3160 and 3180. Compare values For example, if the addition value for the row is smaller than the addition value for the column, the vertical orthogonal transformation is performed after the transverse orthogonal transformation, and the inverse orthogonal transformation is performed inversely to the orthogonal transformation, while the addition value for the row is the addition value for the column. If larger, a first comparator 3190 for outputting a conversion direction control signal to reverse the process and orthogonal transformation and inverse orthogonal transformation is further provided.

In this case, the horizontal subtractor 3110 and the vertical subtractor 3130 selectively receive shape information input from the additional information encoder 9, and when the shape information is input, the pixel value existing in the part to be encoded, that is, the object part. Use only these to find the difference between pixel values.

Fig. 14 is a diagram showing Embodiment 2 of mode control means, in which the mode control means according to Embodiment 2 is a shape information signal selectively inputted from the additional information encoder 9 and a current inputted from the subtractor 1; An orthogonal transformer 5 (3200) for receiving a prediction error signal block to be encoded, and then transversally orthogonally transform the vertical orthogonal transform, and outputs a transform coefficient corresponding to the orthogonal transformer. A first bit stream counter 3220 for quantizing the variable length and checking a code length generated by the variable length coding, a shape information signal selectively input from the additional information encoder 9, and a prediction error signal from the subtractor 1; Orthogonal transformer 6 (3210) and the orthogonal transformer 6 (3210) for outputting the transform coefficient corresponding to the transverse orthogonal transformation after the vertical orthogonal transformation to receive the block, the transform coefficient is input from the orthogonal transformer 6 (3210) The second bit stream counter 3230, which checks the code amount generated when the transform coefficient is quantized and variable length encoded, is compared with the code amounts calculated by the first and second bit stream counters 3220 and 3230. And a second comparator 3240 for outputting a conversion direction control signal for performing orthogonal transformation for generating a small amount of code. In this case, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory 7 as well as the prediction error between the signal predicted from the first memory 7 and the current input signal. To this end, the value stored in the first memory may be initialized to 0 to make the same as the signal from the subtractor and the input image signal.

Fig. 15 shows a third embodiment of the mode control means, in which the mode control means according to the third embodiment is a shape information signal selectively inputted from the additional information encoder 9 and a current inputted from the subtractor 1; An orthogonal transformer 7 (3300) for receiving the prediction error signal block to be encoded and performing vertical orthogonal transformation and transverse orthogonal transformation, and outputting a corresponding transform coefficient; and a quantizing unit for receiving the transform coefficient from the orthogonal transformer 7 (3300). A fifth quantizer 3310, a fifth inverse quantizer 3320 for receiving and inversely quantizing a transform coefficient block quantized by the fifth quantizer 3310, and optionally input by the side information encoder 9 Inverted orthogonal transformer 5 (3330) for receiving the shape information signal and the inverse quantized transform coefficient from the fifth inverse quantizer 3320 and transversing the inverse orthogonal transformation and then outputting the reconstructed image signal block by performing the vertical inverse orthogonal transformation. And a first error for receiving an error signal of the two blocks by receiving the image signal block reconstructed by the inverse orthogonal transformer 5 (3330) while receiving the prediction error signal block to be encoded which is input to the quadrature transformer 7 (3300). The measuring device 3340 is provided. In this case, the prediction error signal block may be initialized to 0 to be equal to the input image signal block.

In addition, the mode control means according to the third embodiment receives an prediction error signal block to be currently encoded from the subtractor 1, performs transverse orthogonal transformation, and then performs vertical orthogonal transformation to output a corresponding transform coefficient. A third quantizer 3360 for receiving and quantizing a transform coefficient from the quadrature converter 8 (3350), and an inverse quantizing unit for receiving a quantized transform coefficient block from the sixth quantizer 3360 (3350); An inverse transformer 6 that receives an inverse quantized transform coefficient from the six inverse quantizer 3370 and the sixth inverse quantizer 3370, outputs a reconstructed image signal block by performing vertical inverse orthogonal transformation and horizontal inverse orthogonal transformation. 3380 and the prediction error signal block to be encoded which are input to the quadrature converter 8 (3350) and the video signal block reconstructed by the inverse orthogonal transformer 6 (3380) to generate an error between the two blocks. Orthogonal and inverse transforms are performed on an orthogonal transformer used for a node having a small error by comparing the error calculated by the second error measuring device 3390 and the first error measuring device 3340 and the second error measuring device 3390. A third comparator 3400 for outputting a conversion direction control signal for performing orthogonal transformation is further configured. At this time, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory as well as the prediction error between the signal predicted from the first memory and the current input signal. The stored value can be initialized to 0 to make it equal to the signal from the subtractor and the input video signal.

An encoding method and a decoding method using an image compression encoding and decoding apparatus using the adaptive conversion method according to the second embodiment of the present invention configured as described above will be described.

First, the coding method will be described.

The image signal block is input to the subtractor 1, and the shape information is input to the additional information encoder 9 and encoded. At this time, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory as well as the prediction error between the signal predicted from the first memory and the current input signal. The stored value can be initialized to 0 to make it equal to the signal from the subtractor and the input video signal.

At this time, the subtractor 1 subtracts the input image signal block and the reconstructed image signal block output from the first memory 7 to obtain a prediction error signal block, and outputs the signal to the orthogonal transformation means 1000.

On the other hand, the mode control means 3000 receives the prediction error signal block from the subtractor 1 and outputs a conversion direction control signal for determining the orthogonal transformation and the inverse orthogonal transformation direction order by using the same, and the control signal encoder 4000 The conversion direction control signal is received and encoded and output to the multiplexer 11.

The orthogonal transformation means 1000 receives the prediction error signal block from the subtractor 1, and selectively receives shape information from the additional information encoder 9, and the orthogonal transformation direction is changed by the mode control means 3000. It is controlled to output a corresponding conversion coefficient signal.

Thereafter, the first quantizer 3 converts the transform coefficient output from the orthogonal transform means 1000 into a quantized transform coefficient block, and the variable length coder 8 is quantized by the first quantizer 3. A transform coefficient block is input and converted into a variable length coded signal.

The multiplexer 11 receives the variable length coded signal from the variable length encoder 8 and selectively receives the shape information signal of the variable length coded from the side information encoder 9 and simultaneously controls the control signal encoder ( The conversion direction control signal encoded at 4000 is received and the signals are multiplexed and transmitted through the transmission medium 12.

Meanwhile, the quantized transform coefficient block output from the first quantizer 3 is converted into an inverse quantized transform coefficient block which is input to the first inverse quantizer 4 and the first inverse orthogonal transform means 140 The first inverse quantizer 4 receives an inverse quantized transform coefficient block, and selectively receives shape information encoded by the side information encoder 9 by inverse orthogonal transformation by the mode control means 3000. The direction is controlled to output the corresponding inverse transform coefficient.

Subsequently, the first adder 6a receives the inverse transform coefficient from the first inverse orthogonal transformation means 140 and simultaneously receives the image signal block previously restored from the first memory 7 and adds the two signals to the original. The video signal block is restored, and the video signal block is stored in the first memory 7, and the first memory 7 outputs the restored video signal block to the subtractor 1.

And, the method for generating the conversion direction control signal in the mode control means 3000 will be described for each embodiment described above.

A method of generating a conversion direction control signal according to Embodiment 1 will be described with reference to FIG. 13. First, the 2D memory 3100 receives and stores a prediction error signal block to be currently encoded by the subtractor 1, and then , 2 switches 3151-a and 3115-b are turned off with the first comparator, and the values of the first and second temporary memories 3160 and 3180 are initialized to zero.

Subsequently, the prediction error signal block stored in the 2D memory 3100 is horizontally subtracted by the horizontal and vertical subtractors 3110 and 3130, and then subtracted between pixel values in the row and column directions. And the vertical adders 3120 and 3140, respectively, to obtain the sum of their magnitude values.

The sum of the size values is transferred to the final horizontal and vertical adders 3150 and 3170, respectively, and is added to the sum of the previous size values stored in the first and second temporary memories 3160 and 3180, respectively. 3160 and 3180, respectively.

For the next row and column of the prediction error signal block, the sum of the error values between the pixels in all the row and column directions is calculated using the method described above, and finally the first and second switches are turned on for the first comparator. The outputs of the final horizontal and vertical adders 3150 and 3170 are transmitted to the first comparator 3190 so that the first comparator 3190 compares the sum of the error values between the pixels in the row and column directions with each other and adds up the error values. It is determined that the similarity of the small direction is large, and the first direction is orthogonally transformed into the direction of the similarity, and then orthogonally transformed to the other direction. Output

And, the method of generating the conversion direction control signal according to the second embodiment will be described with reference to FIG. 14. First, when the prediction error signal block is input to the apparatus from the subtractor 1, the quadrature converter 5 (3200) is the prediction error. After the transverse orthogonal transformation of the signal block, the orthogonal transformer 6 3210 transversely orthogonally transforms the prediction error signal block after the vertical orthogonal transformation.

Thereafter, the first and second bitstream counters 3220 and 3230 receive the conversion coefficients obtained by the orthogonal transformation in the quadrature converters 5 and 6 (3200 and 3210), respectively, and calculate a code generation amount.

Subsequently, the second comparator 3240 compares the code generation amounts calculated by the first and second bitstream counters 3220 and 3230 with each other so as to perform an orthogonal transformation performed by an orthogonal transformer having a small code generation direction. The orthogonal transformation outputs a conversion direction control signal for performing in the opposite direction to the orthogonal transformation.

In addition, the method of generating the conversion direction control signal according to the third embodiment will be described with reference to FIG. 15. First, when the prediction error signal block is input to the apparatus from the subtractor 1, the quadrature converter 7 (3300) performs the prediction error. After transverse orthogonal transformation of the signal block, transverse orthogonal transformation is performed. At this time, the subtractor 1 may output the current input signal as it is without prediction from the signal of the first memory as well as the prediction error between the signal predicted from the first memory and the current input signal. The stored value can be initialized to 0 to make it equal to the signal from the subtractor and the input video signal.

First, if the prediction error signal block is input to the apparatus in the subtractor 1, the orthogonal transformation

Subsequently, the fifth quantizer 3310 receives and quantizes the transform coefficient from the quadrature transformer 7 (3300), and the fifth inverse quantizer 3320 receives the quantized transform coefficient block and dequantizes it. The converter 53330 receives the inverse quantized transform coefficient block and performs inverse orthogonal transformation in a direction opposite to the conversion direction of the quadrature transformer 7 (3300) to output a restored signal block.

The first error measurer 3340 calculates an error between the signal block reconstructed by the inverse orthogonal transformer 5 3330 and the predicted error signal block input to the orthogonal transformer 7 3300 to the third comparator 3400. The error value is output.

On the other hand, orthogonal transformer 8 (3350) existing in another node transversely orthogonally transforms the prediction error signal block and then orthogonally transforms it, and the sixth quantizer 3360 inputs a transform coefficient from the orthogonal transformer 8 (3350). And a sixth inverse quantizer 3370 receives and inverse quantizes the quantized transform coefficient block. An inverse orthogonal transformer 6 3380 receives the inverse quantized transform coefficient block and receives the inverse quantized transform coefficient block 8 (3350). Inverted orthogonal transformation is performed in a direction opposite to the transformation direction of the output signal).

The second error measurer 3390 calculates an error between the signal block reconstructed by the inverse orthogonal transformer 6 (3380) and the predicted error signal block input to the orthogonal transformer 8 (3350) to the third comparator 3400. The error value is output.

In this case, the third comparator 3400 receives the error values from the first and second error measuring units 3340 and 3390, respectively, and compares the error values, and is performed by the quadrature and inverse transformers of the node having the smallest error value. Outputs a conversion direction control signal for conversion and inverse orthogonal conversion.

Meanwhile, the decoding method will be described.

First, the demux unit 13 receives a coded signal through the transmission medium 12 and demultiplexes it to output variable length coded shape information, an image signal, and a conversion direction control signal. The coded variable length image signal is received and converted into a quantized transform coefficient signal, while the additional information decoder 14 selectively receives the variable length coded shape information and restores the shape information.

The control signal decoder 5000 receives the encoded conversion direction control signal and restores the original conversion direction control signal.

In this case, the second inverse quantizer 200 receives the quantized transform coefficient signal from the decoder 15 and converts it into an inverse quantized transform coefficient block, and the second inverse orthogonal transform means 2000 performs the second inverse quantization. While receiving the inverse quantized conversion coefficient from the device 200, selectively receives the shape information restored from the additional information decoder 14, and receives the conversion direction control signal restored from the control signal decoder (5000) By performing inverse orthogonal transformation in the direction opposite to the orthogonal transformation direction of the orthogonal transformation means 1000, the inverse orthogonally transformed signal block is output.

Subsequently, the second adder 6b receives the signal block de-orthogonally transformed by the second inverse orthogonal transformation means 2000 and the image signal block previously restored in the second memory 18, and adds the two signal blocks. To restore the original video signal block.

Meanwhile, the second memory 18 receives and stores the reconstructed video signal block output from the second adder 6b and outputs the reconstructed video signal block to the second adder 6b.

When the second inverse orthogonal transformation means 200 performs an inverse orthogonal transformation by the conversion direction control signal output from the control signal decoder 5000, the conversion direction control signal after the transverse orthogonal transformation When the vertical orthogonal transformation information is included, the second inverse orthogonal transformation means 200 performs horizontal inverse orthogonal transformation after vertical inverse orthogonal transformation, and when the conversion direction control signal has transverse orthogonal transformation information after vertical orthogonal transformation, After the inverse orthogonal transformation, longitudinal inverse orthogonal transformation is performed.

The use of the mode controller to set the conversion direction may be limited by the characteristics of the signal or the complexity of the hardware implementation. As an example of the former, when the signal components are very uniformly distributed, only a small gain or no gain for the coding efficiency obtained as a change in the transformation direction can be obtained. In this case, additional code generation due to transmission of the control signal lowers the overall coding efficiency. As the latter example, in the case of a portable terminal, the complexity of hardware implementation may be a problem greater than the coding efficiency, and there is a case where the hardware implementation should be simplified at the expense of an increase in the coding efficiency that can be obtained by adaptive setting of the conversion direction. However, in terms of compatibility of the bit strings, a coder that does not adaptively set the transform direction should also be able to decode the code string coded by a decoder that includes a control signal for the transform direction. To this end, information indicating whether the control signal for the conversion direction is used must be included in the bit string.

FIG. 16 is a control block diagram of an image compression encoding and decoding apparatus using an adaptive transform method according to a third embodiment of the present invention for limiting the use of a mode controller as described above. And a mode controller control signal generator 6000, a mode controller control signal encoder 7000, and a mode controller control signal decoder 8000 are added to the decoding apparatus.

The mode controller control signal generating means 6000 controls the mode control means 3000 according to whether or not the orthogonal conversion means 1000 and the first and second inverse orthogonal conversion means 140 use the conversion direction control signal. It serves to generate a mode controller control signal for controlling the on / off operation of the signal encoder 4000 and the control signal decoder 5000.

The on / off operation of the mode control means 3000, the control signal encoder 4000 and the control signal decoder 5000 is equally applied according to the mode controller control signal generated by the mode controller control signal generating means for the encoding target. .

The mode controller control signal encoder 4000 receives and encodes a mode controller control signal from the mode controller control signal generating means 6000 and inputs the multiplexer 11 to multiplex with other signals.

The mode controller control signal decoder 8000 receives and decodes the coded mode controller control signal through the multiplexer 11, the transmission medium 12, and the demux 13, and decodes the decoded mode controller control signal. As a function of controlling the operation of the control signal decoder 5000.

An encoding device and a decoding method of the encoding device and the decoding device of FIG. 16 having the above-described configuration will be described.

The operation of the apparatus is determined by the use of the conversion direction control signals of the orthogonal transformation means 1000 and the first and second inverse orthogonal transformation means 140 and 200.

The transmission of the mode controller control signal, which is information on whether or not to use the conversion direction control signal for setting the conversion direction, is performed by the start position of one MB (macro-block), the start position of one screen, the start position of a scene, or the entire video. Includes information on whether to use the control signal for setting the conversion direction at the start position. Although the code amount is increased due to the transmission of the mode controller control signal, the increase in the code amount due to the transmission of the mode controller control signal is negligible compared to the overall code amount.

If the conversion direction control signal is used, the mode controller control signal generating means 6000 outputs a mode controller control signal corresponding thereto, so that both the mode control means 3000 and the control signal encoder 4000 are used. In this case, the mode controller control signal encoder 7000 encodes the mode controller control signal, and the encoded control signal is transmitted through the multiplexer 11, the transmission medium 12, and the demux unit 13. The mode controller control signal decoder 8000 is input and decoded, and the decoded mode controller control signal operates the control signal decoder 5000.

Subsequently, the encoding and decoding process is the same as the above-described encoding method and decoding method of FIG. 12.

On the other hand, when the conversion direction control signal is not used, the mode controller control signal generating means 6000 outputs a mode controller control signal corresponding thereto, and accordingly, the operation of the mode control means 3000 and the control signal encoder 4000 is performed. The mode controller control signal encoder 7000 encodes the mode controller control signal, and the encoded control signal is controlled by the multiplexer 11, the transmission medium 12, and the demux unit 13. The decoded mode controller control signal is inputted to the signal decoder 8000 and the operation of the control signal decoder 5000 is stopped.

When the conversion direction control signal is not used, the orthogonal transformation means 1000 performs a predetermined conversion method for this case among the transverse orthogonal transformation and the transverse orthogonal transformation after transversal orthogonal transformation of the prediction error signal block for encoding. Perform. In addition, the inverse orthogonal transformation means performs an inverse transformation method corresponding to the orthogonal transformation means 1000. That is, when the orthogonal transform means 1000 transversely orthogonally transforms the predicted error signal block input after the vertical orthogonal transformation when the conversion direction control signal is not used, the inverse orthogonal transform means crosses the input dequantized transform coefficient block. After inverse orthogonal transformation, vertical inverse orthogonal transformation is performed to output an inverse orthogonal transformation signal block.

In this case, the inverse quantized transform coefficient block input through the second inverse quantizer 200 is inversely orthogonally transformed by the second inverse orthogonal transform means 2000 in the direction opposite to the orthogonal transform direction of the orthogonal transform means 1000. The rest of the process is the same as the decoding process of FIG.

On the other hand, the method of configuring the mode controller can be used in addition to the above-described embodiments, the method is a method using the average value change rate of the surrounding blocks including the current block, respectively projected horizontally and vertically to the coding block By using the rate of change of the projected values, the method of determining the transformation method using only a part of the block to be encoded, such as the peripheral value of the block to be encoded or a part of the center, can be used.

That is, by additionally transmitting the conversion direction control signal, the conversion method can be very flexibly determined.

Determination of the conversion method by transmitting the conversion direction control signal is performed only in the encoder and the decoder only changes the conversion method according to the transmitted control signal, thereby increasing the complexity of the decoder.

As described above, according to the video compression encoding and decoding apparatus using the adaptive transformation method according to the present invention and the method, the information of the coded block, or the information of the coded block and information about the block to be currently encoded Whether the signals in the block have a high correlation in the vertical direction or a high correlation in the horizontal direction is judged, and orthogonal transformation is first performed in the direction of high correlation, and then the orthogonal transformation is performed in the direction of low correlation. On the other hand, the compression encoding efficiency can be improved by performing inverse orthogonal transformation as opposed to the orthogonal transformation direction.

The method of determining the transform direction of the block to be encoded using the current block to be encoded is more efficient than the method of determining the transform direction using the encoded block. appear. In this case, the editability of the code string refers to the ability to easily change an image property such as texture or color of the image by manipulating the code string.

That is, the ability to replace the bit string for the part to change the attribute in the image with a new bit string reflecting the changed attribute, and the other part is to use the bit string already generated and encoded as it is. In the case of not transmitting the information about the uplink, the bit streams generated are not independent of each other and are previously spatially applicable because the information about the already encoded part is continuously used to obtain a transform direction for the current encoding part. It continues to relate to the bit strings that become. Therefore, if you change the property of any part of the image, there is a disadvantage that you need to regenerate the code strings for all the signals after the part where the property is changed spatially. However, in case of additionally transmitting information on the transformation direction, the bit stream for the signal part to be encoded and the already encoded signal parts can be generated independently, and the property is changed when the property of one part in the signal is changed. By re-coding only the bit stream for the part and replacing the bit stream belonging to the same position of the existing bit stream, the effect of changing only the attributes of a part of the signal is shown.

1 is a control block diagram of a video compression encoding and decoding apparatus according to the prior art;

2A to 2E are diagrams illustrating a process of orthogonal transformation using shape information in the orthogonal transformer of FIG. 1;

3A to 3C are diagrams for explaining that conversion coefficients are different according to a direction orthogonal transformation in the orthogonal transformer of FIG. 1;

4 is a control block diagram of an image compression encoding and decoding apparatus using an adaptive transform method according to an embodiment of the present invention;

FIG. 5 is a control block diagram of orthogonal transform means and first and second inverse orthogonal transform means of FIG. 4;

6 is a control block diagram of orthogonal transform means and first and second inverse orthogonal transform means according to another embodiment of FIG. 4;

7 shows a first embodiment of a mode controller;

8 shows Embodiment 2 of a mode controller.

9 is a view showing an input image having star shape information;

FIG. 10 illustrates a block for encoding an input image having the shape information of FIG. 9; FIG.

FIG. 11 is a diagram illustrating an example in which a transform method for a signal to be encoded cannot be selected because no neighboring information exists when the shape information of FIG. 9 is used and the block unit coding method is used.

12 is a control block diagram of an image compression encoding and decoding apparatus using an adaptive transform method according to a second embodiment of the present invention;

FIG. 13 shows an embodiment 1 of the mode control means of FIG. 12;

14 shows a second embodiment of the mode control means of FIG.

15 shows a third embodiment of the mode control means of FIG.

16 is a control block diagram of an image compression encoding and decoding apparatus using an adaptive transform method according to a third embodiment of the present invention.

<Description of Symbols for Main Parts of Drawings>

100: conversion mode control means 140: first inverse orthogonal conversion means

210: reverse conversion mode control means 1000: orthogonal conversion means

2000: second orthogonal transformation means 3000: mode control means

4000: Control signal decoder 5000: Control signal decoder

Claims (6)

A coding apparatus for compressing an input video signal block by orthogonal transform, Orthogonal transform of the input image signal block into a highly correlated direction by selectively receiving the transform coefficient obtained by the orthogonal transformation and the video signal block to be currently encoded, and determining the correlation between the input values in the horizontal and vertical directions. A conversion mode control means for outputting a conversion direction control signal for performing orthogonal conversion in a direction different from that; And an orthogonal transform means for receiving a transform direction control signal from the transform mode control means and performing orthogonal transform corresponding thereto. The method of claim 1, The conversion mode control means includes a memory for storing a conversion coefficient obtained after orthogonal conversion; And a mode controller for receiving a transform coefficient from the memory and selectively receiving an image signal block to be encoded, and outputting a transform direction control signal for controlling the transform direction order of the orthogonal transform means. Image compression coding device. The method of claim 2, The mode controller obtains a block average value of an already orthogonally transformed transform coefficient around a video signal block to be encoded, obtains a rate of change in the longitudinal direction and a rate of change in the transverse direction, and compares the absolute values of the two rate of change with each other. And an orthogonal transformation of an input image signal block by first outputting a value for a small direction, and then performing orthogonal transformation in a direction different from that of a small change rate. The method of claim 2, The mode controller obtains a block average value of the video signal block to be encoded currently and the orthogonal transform coefficients in the periphery thereof, obtains the vertical change rate and the horizontal change rate of the average values, and stores the absolute values of the change rate in the vertical and horizontal directions. And then orthogonally transform the input image signal block into a smaller direction by comparing the magnitudes of the sums of the two directions and outputting the value for the smaller direction. And a video compression encoding apparatus using an adaptive transformation method. The method of claim 1, The orthogonal transform means receives an image signal block to be encoded, selectively receives shape information, and orthogonally transforms the image signal block to be encoded in a vertical direction, and then orthogonally transforms in a horizontal direction; An orthogonal transformer 2 configured to receive the image signal block to be encoded, selectively receive shape information, and orthogonally transform the image signal block to be encoded in a horizontal direction and then orthogonally transform it in a vertical direction; A first switch receiving a conversion direction control signal from the conversion mode control means and switching to intermittently input the video signal block to be encoded into the quadrature converter 1 or the quadrature converter 2; And a second switch which receives a conversion direction control signal from the conversion mode control means and switches to control the output of the conversion coefficient from the quadrature converter 1 or the quadrature converter 2. Image compression coding device. The method of claim 1, The orthogonal transformation means includes a preprocessor which receives a conversion direction control signal from the conversion mode control means and receives an image signal block to be encoded in an on / off mode and rotates it 90 degrees clockwise; An orthogonal converter receiving an image signal block rotated by 90 degrees from the preprocessor, receiving oral shape information, and orthogonally transforming in a vertical / horizontal direction and then orthogonally transforming in a horizontal / vertical direction; The conversion mode control means receives a conversion direction control signal and the on / off mode is controlled in the on mode, the conversion coefficient is received from the quadrature converter rotates 180 degrees around the axis connecting the upper left and lower right of the two-dimensional block An image compression encoding apparatus using an adaptive conversion method comprising a post processor.