JP2004206725A - Method and device for dividing and synthesizing sub-band - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a high speed sub-band division device with satisfactory conversion efficiency. <P>SOLUTION: An inputted image signal s10 is connected with a sub-band division filter 11 which performs sub-band division in the horizontal direction at the first stage, output s11-1 of low frequency components of the sub-band division filter is connected with a sub-band division filter 12 which performs sub-band division in the horizontal direction at the next stage and on the other hand, output s11-2 of high frequency components of the sub-band division filter 11 is connected with a memory 14 which stores a sub-band division result in the horizontal direction. Output of a sub-band division filter at the second stage is also performed similarly as that of the sub-band division filer at the first stage. Thus, the sub-band division in the horizontal direction of n stages is performed by cascading n stages of sub-band division filers. In addition, output of the memory 14 is connected with a sub-band division filer 15 which performs sub-band division in the vertical direction at the first stage, processing similar to the sub-band division in the horizontal direction is performed and the high frequency components are outputted to the outside. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

  The present invention relates to an image encoding / decoding apparatus for a still image or a moving image, and relates to an efficient image encoding.

  2. Description of the Related Art Conventionally, a wavelet transform of a two-dimensional image is configured by connecting wavelet filters that perform one-dimensional wavelet transform in multiple stages (for example, see Non-Patent Document 1).

  The wavelet filter includes a pair of a low-pass filter 32-1 and a high-pass filter 32-2 as shown in FIG. In FIG. 3, when the one-dimensional signal S31 is input, the low-pass filter 32-1 applies a low-pass filter to the input signal as shown in Expression 1, sub-samples the output data by half, and outputs the data. . The high-pass filter 32-2 applies a high-pass filter to the input signal as shown in Expression 2, sub-samples the output data by half, and outputs the result (S32-2). In Equations 1 and 2, h (k) and g (k) represent coefficients of a low-pass filter and a high-pass filter, respectively, S1 represents an input signal, and Sl + 1 and Wl + 1 represent a low-pass filter and a high-pass filter, respectively. The output of

  FIG. 2 is a configuration diagram of the wavelet transform described in Non-Patent Document 1. Hereinafter, the configuration of the conventional wavelet transform will be described with reference to FIG.

In FIG. 2, a two-dimensional image signal S20 sequentially input one-dimensionally in a horizontal direction (or a vertical direction) is first connected to a wavelet filter 21 for performing a one-dimensional wavelet transform in a first stage, and outputs a low-frequency component thereof. S21-1 and the output S21-2 of the high frequency component are stored in the first memory 22. After completion of the first-stage one-dimensional wavelet transform, the low-frequency component S22-1 and the high-frequency component S22-2 of the image sequentially output one-dimensionally in the vertical (or horizontal) direction from the first memory are respectively converted to the second. It is input to wavelet filters 23-1 and 23-2 for performing one-dimensional wavelet transform in the first stage, and as a result of the wavelet transform, low-frequency components S23-1 both vertically and horizontally and low-frequency components S23-2 and S23-only only vertically or horizontally. 3 and high-frequency components S23-4 in both the vertical and horizontal directions are respectively stored in the second memory. Thus, one-stage two-dimensional wavelet transform is completed. When the low-frequency component S23-1 in both the vertical and horizontal directions is input to the first-stage wavelet filter 21, the two-dimensional wavelet transform is repeatedly performed on the low-frequency component in both the vertical and horizontal directions. For example, frequency band division of an image by three-dimensional two-dimensional wavelet transform is as shown in FIG.
Stephane Mallat A Theory for Multiresolution Signal Decomposition: The Wavelet Representation, IEEE Trans. PAMI, Vol. 11, No. 7, July 1989. pp. 674-693.

  However, the conventional wavelet transform apparatus has the following problems. First, in the conventional wavelet transform device, the next-stage transform cannot be performed until the two-dimensional wavelet transform in the preceding stage is completed, so that much time is required for the transform and it is not suitable for speeding up. Each frequency band after conversion is as shown in FIG. 4. For example, in a band such as HL3 in FIG. 4, when viewed from the horizontal direction, it is a high frequency band, but in a vertical direction, it is a low frequency band. Therefore, a high correlation is still left in the vertical direction, which causes a decrease in the conversion compression efficiency. Further, in the conventional wavelet transform device, the input image needs to be rectangular. Therefore, an image of an arbitrary shape cannot be converted as it is. In order to perform the conversion, it is necessary to fill a blank area without image data in the image. As a generally conceivable method of filling, for example, a method of filling the average value of the image area or a method of copying and filling the pixel at the end of the image area can be considered. In many cases, a large gap is formed, or a large amount of high-frequency components are generated in a blank area, which greatly affects the subsequent coding efficiency. Further, the conventional wavelet transform method is suitable for a local image change, but is not suitable for a periodically changing image. This is because the wavelet transform roughly divides high frequency components. Further, in the conventional wavelet transform method, there is a disadvantage that a large number of multiplications are required in the calculation of the wavelet filter, which requires a long calculation time and a heavy load on the device.

  In order to solve the above-mentioned problems, the problems are solved by the following devices, a method that provides the same effect as those devices, or a combination of these devices and methods.

  A conversion control means for outputting to the selector a control signal for inputting or determining whether or not to perform a predetermined sub-band division; a two-dimensional image signal or a two-dimensional image signal sequentially input in the horizontal direction (or the vertical direction); One of signals in a predetermined frequency band of a two-dimensional image sequentially and one-dimensionally input in a horizontal direction (or a vertical direction) from a memory is selected by a control signal output from the conversion control means, and a sub signal is selected. A band division filter or a selector for outputting to the memory, and a signal output from the selector is subjected to one-dimensional sub-band division and divided into two frequency bands in a horizontal direction (or a vertical direction) to output a signal. Before storing and outputting the sub-band splitting filter, the signal output from the sub-band splitting filter, and the signal output from the selector. It includes a memory, a.

  Or a sub-band division filter comprising a rational number or a product of a rational number and a square root of 2 or a filter coefficient having an approximate value thereof, wherein the sub-band division operation is performed by addition and subtraction and shift operation and one multiplication or division This is performed using

  Inverting control means for outputting to the selector a control signal as to whether or not to perform input or predetermined sub-band synthesis, or output from an input sub-band divided image signal or sub-band synthesis filter A memory for storing a signal to be output or a signal output from the selector; and a control signal output from the inverse conversion control means, the memory storing the signal in a horizontal (or vertical) direction for each predetermined frequency band. A selector that outputs a sequentially input two-dimensional image signal to a sub-band synthesis filter or the memory; and a one-dimensional sub-band synthesis is performed on the signal output from the selector to restore an image and output a signal. And a sub-band synthesis filter.

  Or a subband synthesis filter comprising a rational number or a product of a rational number and the square root of 2 or a filter coefficient having an approximate value thereof, wherein the subband synthesis operation is added and subtracted and shifted, and one multiplication or division is performed. A subband synthesis filter characterized by performing the following.

<< Specific Example 1 >>
<Configuration> Hereinafter, a first embodiment of the present invention will be described in detail with reference to the drawings. FIG. 1 is a configuration diagram of the subband division of the first embodiment according to the present invention. The subband splitting device according to the first embodiment of the present invention includes three stages of subband splitting filters for performing one-dimensional subband splitting in the horizontal direction, a memory for storing a subband splitting output in the horizontal direction, It has three stages of sub-band division filters for performing one-dimensional sub-band division.

  In FIG. 1, an input image signal s10 is connected to a sub-band division filter 11 that performs a first-stage horizontal sub-band division, and an output s11-1 of a low-frequency component of the sub-band division filter is output to a next stage. Is connected to a sub-band division filter 12 that performs horizontal sub-band division, while the output s11-2 of the high-frequency component of the sub-band division filter 11 is connected to a memory 14 that stores a horizontal sub-band division result. Have been. Similarly to the output of the first-stage sub-band splitting filter, the low-frequency component of the output of the second-stage sub-band splitting filter is connected to the next-stage sub-band splitting filter, and the high-frequency component is stored in the memory. In this way, the three-stage sub-band division filters are cascaded, and three-stage horizontal sub-band division is performed. Further, the output of the memory 14 is connected to a sub-band division filter 15 for performing vertical sub-band division of the first stage, and the output s15-1 of the low-frequency component of the sub-band division filter is output to the next stage of the vertical stage. It is connected to a sub-band division filter 16 that performs sub-band division in the direction, while the output s15-2 of the high-frequency component is output to the outside. Similarly to the output of the first-stage sub-band division filter, the low-frequency component of the output of the second-stage sub-band division filter is connected to the next-stage sub-band division filter, and the high-frequency component is output to the outside.

  <Operation> When an image signal is input, the first-stage subband division filter 11 in the horizontal direction performs one-dimensional subband division in the horizontal direction as shown in Expressions 1 and 2, and outputs a low-frequency component. s11-1 is output to the next-stage sub-band division filter, and the high-frequency component s11-2 is output to the memory 14. The sub-band division filters at each stage in the horizontal direction operate in the same manner as the sub-band division filter at the first stage, perform sub-band division in the horizontal direction one after another, and store the output in the memory 14. In this way, three horizontal sub-band divisions can be performed continuously. After the horizontal sub-band division is completed, the image data in the vertical direction is output from the memory 14 and is used as an input for the vertical sub-band division. The vertical sub-band division has the same configuration and the same operation as the horizontal sub-band division, and performs three vertical sub-band divisions. In this manner, two-dimensional three-stage subband division can be performed by one memory storage process, and high-speed operation is possible.

  In addition, since the signal thus divided into sub-bands is divided into frequency bands as shown in FIG. 5, as compared with FIG. 4, the low-frequency region is finely divided in both the vertical and horizontal directions. Correlation within the frequency band is reduced, and compression coding efficiency is improved.

  <Effects> As described above, according to the first embodiment of the present invention, two-dimensional three-stage subband division can be performed by one memory storage process, and high-speed operation is possible. In addition, since the low-frequency region of the sub-band divided signal is finely divided in the vertical and horizontal directions as shown in FIG. 5, the correlation in each frequency band is reduced, and the compression encoding efficiency is improved. .

<< Specific Example 2 >>
<Structure> Hereinafter, a second embodiment of the present invention will be described in detail with reference to the drawings. The image coding apparatus according to the second embodiment of the present invention is the same as the image coding apparatus according to the first embodiment of the present invention, except that a filter is determined in advance so that transform coefficients of three-dimensional one-dimensional subband division can be obtained. Is calculated in advance, and the three-stage sub-band division is replaced with a one-stage matrix operation.

  FIG. 6 is a block diagram of an image coding apparatus according to a second specific example of the present invention. In FIG. 6, an input image signal is connected to a matrix operation means 41 which performs a first-stage three-stage subband division in a horizontal direction, an output s41 of which is connected to a memory 42, and an output s42 of the memory 42. Is connected to a matrix operation means 43 for performing a three-stage vertical sub-band division in the second stage at a time, and an output s43 thereof is connected to the outside.

  <Operation> In the sub-band dividing apparatus according to the first embodiment of the present invention shown in FIG. 1, three-dimensional one-dimensional sub-band division can be represented by three matrix operations as shown in Expression 3.

  In Equation 3, x and y represent the input signal and the output of the one-dimensional subband split, h and g represent the coefficients of the low-pass and high-pass filters, Φ represents a matrix with all coefficients zero, and I is the unit. Represents a matrix. Here, as an example, a filter coefficient of two taps and subband division of eight data are shown, but a filter of an arbitrary tap and subband division of an arbitrary number of data can be similarly expressed. If the matrix of the three sub-band division filter coefficients of Expression 3 is calculated in advance, the operation of Expression 3 can be performed by one matrix operation as in Expression 4. Therefore, the first embodiment of the present invention shown in FIG. 1 first performs a matrix operation on data in the horizontal direction as shown in FIG. 6, stores the output in a memory, and outputs the data vertically from the memory. By performing a matrix operation on the vertical data, the same subband division as in the first embodiment of the present invention can be realized. By using this method, it is possible to realize a higher speed with a simpler device.

  <Effects> As described above, according to the second embodiment of the present invention, in the first embodiment of the present invention, the three-stage sub-band division is performed by one-stage matrix operation, thereby making it easier. With a simple device.

  In the first and second specific examples described above, the configuration is such that the vertical sub-band division is performed following the horizontal sub-band division, but the present invention is not limited to this. . For example, the vertical sub-band division may be performed first, and then the horizontal sub-band division may be performed, and a similar effect may be obtained.

<< Specific Example 3 >>
<Structure> Hereinafter, a third embodiment of the present invention will be described in detail with reference to the drawings. FIG. 7 is a configuration diagram of the shape adaptive subband division of the third embodiment according to the present invention. A shape adaptive sub-band dividing apparatus according to a first specific example of the present invention is a conventional sub-band dividing apparatus which enables input of shape information, divides only an image area indicated by the shape information into sub-bands, and The information also has a shape changing means so as to indicate the output position of the subband division.

  In FIG. 7, the shape information s1-1 is connected to one of the horizontal shape changing means 2-1 and the sub-band dividing means 2-2, and the output s2-1 of the shape changing means 2-1 is stored in the memory means 3 Connected to -1. On the other hand, the input image signal s1-2 is connected to the other of the horizontal sub-band dividing means 2-2, and the output s2-2 of the low frequency component of the sub-band dividing means 2-2 is stored in the memory means 3 -2, the output s2-3 of the high-frequency component of the sub-band dividing means 2-2 is connected to the memory means 3-3. The output s3-1 of the memory means 3-1 is connected to one of the vertical shape changing means 4-1 and the sub-band dividing means 4-2 and one of the sub-band dividing means 4-3. The output s4-1 of 4-1 is connected to the memory means 5-1 and the output s3-2 of the memory means 3-2 is connected to the other of the vertical sub-band dividing means 4-2. The output s4-2 of the low-frequency component of the sub-band dividing means 4-2 is connected to the memory means 5-2, and the output s4-3 of the high-frequency component of the sub-band dividing means 4-2 is connected to an external terminal. The output s3-3 of the memory means 3-3 is connected to the other of the vertical sub-band splitting means 4-3, and the output s4-4 of the low frequency component of the sub-band splitting means 4-3 and the output s3-4. Outputs s4-5 of the high-frequency components of the sub-band splitting means 4-3 are respectively connected to external terminals. This completes one-step two-dimensional shape adaptive subband division. Thereafter, in the second and third stages of the two-dimensional shape adaptive subband division, the image information of the band including the lowest frequency component of the previous stage and the shape information of the image are input, and the two-dimensional shape of the first stage is input. The configuration is the same as that of the adaptive subband division.

  <Operation> First, when the two-dimensional image signal s1-2 and the shape information s1-1 of the image signal are input, the horizontal sub-band dividing means 2-2 outputs the horizontal one-dimensional image signal and the According to the corresponding shape information, only the signal in the image area is divided into subbands as in Equations 1 and 2, and is output. For example, assuming that n of the m signals are in the image area, the sub-band dividing means determines (n-1) from the top of the n image signals if n is an odd number. , And the remaining image signal is appended to the end of the (n-1) / 2 outputs of the low-frequency component to obtain (n-1) / 2 + 1 Outputs low frequency components and (n-1) / 2 high frequency components. The shape changing means 2-1 outputs shape information corresponding to the output of the sub-band dividing means. For example, assuming that n of the m signals are in the image area, the shape changing means corresponds to (n-1) / 2 + 1 low frequency components if n is an odd number. The shape information is written from the top in the horizontal direction of the memory means 3-1, and the shape information corresponding to (n-1) / 2 high-frequency components is written from the address m / 2 in the horizontal direction. Thus, the one-dimensional shape adaptive subband division for one line in the horizontal direction is completed. When all of the horizontal one-dimensional subband division is completed, the memory unit 3-1 transfers the shape information of the low-frequency component obtained by the horizontal one-dimensional subband division to the vertical subband division unit 4-2. Then, the shape information of the high-frequency component obtained by the one-dimensional sub-band division in the horizontal direction is output to the sub-band division means 4-3 in the vertical direction. Similarly, the memory means 3-2 and 3-3 also apply the one-dimensional sub-band division of the low frequency component and the high frequency component in the horizontal direction to the sub-band division means 4-2 and the sub-band division means 4- Output to 3. In the sub-band division means 4-2 and the sub-band division means 4-3, respectively, in the same manner as the horizontal sub-band division means 2-2, performs sub-band division according to the input image signal and shape information, The vertical low frequency component and the high frequency component are output. Further, the shape changing means 4-1 also outputs the shape information s4-1 of the position corresponding to the output of the vertical sub-band dividing means in the same manner as the shape changing means 2-1. Store. The output s4-2 of the low-frequency component of the sub-band dividing means 4-2 is also stored in the memory means 5-1. Thus, one-stage two-dimensional subband division is completed. In the sub-band dividing apparatus connected in n stages, the subsequent two-dimensional sub-band dividing unit inputs the low-frequency component of the previous stage and the shape information of the low-frequency component, and performs the same processing as the previous stage.

  <Effects> As described above, according to the third embodiment of the present invention, only signals in an image area are divided into sub-bands by using shape information, and the shape information is also divided into sub-bands. By changing the image signal so as to correspond to the image signal of any shape, it is possible to divide the image signal of an arbitrary shape into subbands at an arbitrary stage. Further, by changing the shape information, the subsequent sub-band division can always be performed on the same frequency component, and sub-band division of an arbitrary shape can be performed without impairing the sub-band division efficiency.

<< Specific Example 4 >>
<Structure> Hereinafter, a fourth embodiment of the present invention will be described in detail with reference to the drawings. FIG. 8 is a configuration diagram of the shape adaptive sub-band division of Example 4 according to the present invention. The shape adaptive sub-band splitting device according to the second embodiment of the present invention includes a shape changing unit n for changing one-dimensional shape information in the horizontal direction, and a sub-band splitting unit n for performing one-dimensional sub-band splitting in the horizontal direction. A stage, a memory for storing the output of the horizontal shape changing means in the last stage and the output of the horizontal subband dividing means, a shape changing means n for changing one-dimensional shape information in the vertical direction, and a vertical direction. Sub-band dividing means for performing one-dimensional sub-band division of n. In FIG. 8, the shape information s3-1 is connected to one of the first horizontal shape changing unit 31-1 and the subband dividing unit 31-2, and the output s31-1 of the shape changing unit 31-1. Are connected to the next-stage shape changing means 32-1 and sub-band dividing means 32-2. The input image signal s3-2 is connected to the other side of the first sub-band dividing unit 31-2 for performing horizontal sub-band division, and the output s31 of the low frequency component of the sub-band dividing unit 31-2. -2 is connected to the next horizontal sub-band division means 32-2, and the output s31-3 of the high frequency component of the sub-band division means 31-2 is a memory for storing the horizontal sub-band division result. Connected to 34-2. The outputs of the second and subsequent shape changing means are connected to the next shape changing means in the same manner as the first-stage output, and the output of the last shape changing means is stored in the memory means 34-1. Is done. In addition, as with the output of the sub-band dividing means of the first stage, low-frequency components are connected to the sub-band dividing means of the next stage, and the outputs of the sub-band dividing means of the second and subsequent stages are also connected. The component and the low frequency component of the last stage are stored in the memory. In this way, the 3n-stage shape changing unit and the 3n-stage sub-band dividing unit are cascade-connected, and sub-band division of the 3n-stage horizontally shaped image is performed. The output of the memory means 34-1 is used as the shape changing means 35-1 for changing the shape information of the first stage in the vertical direction and the sub-band dividing means 35- for performing the first-stage vertical sub-band division. 2, the output s35-1 of the first-stage vertical shape changing unit is connected to the second-stage shape changing unit 36-1 and the second-stage shape changing unit 36-2. I have. The output of the memory means 34-2 is connected to the other of the sub-band dividing means 35-2 for performing the vertical sub-band division of the first stage, and the output of the low-frequency component of the sub-band dividing means 35-2. s35-2 is connected to the next vertical sub-band dividing means 36-2, and the output s35-3 of the high frequency component is output to an external terminal. The outputs of the second and subsequent stages of the shape changing unit are connected to the next stage of the shape changing unit, and the outputs of the last stage of the shape changing unit are connected to the external terminals, similarly to the output of the first stage. Similarly to the output of the sub-band dividing means of the second stage, the low-frequency component is connected to the sub-band dividing means of the next stage, and the high-frequency component of each particular stage and the final stage are output. Is output to the external terminal.

  <Operation> When the two-dimensional image signal s3-2 and the shape information s3-1 of the image signal are input, the first sub-band dividing means 31-2 in the horizontal direction performs the operation of the third specific example. Similarly to the sub-band division means, one-dimensional horizontal sub-band division is performed on only the signal in the image area according to the shape information as in Expressions 1 and 2, and the low-frequency component s31-2 is obtained. The high-frequency component s31-3 is output to the memory 34-2. Similarly to the shape changing means of the third specific example, the first-stage shape changing means 31-1 also converts the shape information of the position corresponding to the output of the sub-band dividing means into the next-stage shape changing means 32-1. Output to -1. The shape changing means and sub-band dividing means at each stage in the horizontal direction perform the same operation as the shape changing means 31-1 and sub-band dividing means 31-2 at the first stage, and successively perform sub-band division in the horizontal direction. Then, the shape information is stored in the memory 34-1 and the subband division output is stored in the memory 34-2. In this way, the subband division of the image signal of an arbitrary shape in the horizontal direction of n stages can be continuously performed. After the horizontal sub-band division is completed, the shape information s34-1 and the image signal s34-2 are output in the vertical direction from the memories 34-1 and 34-2, and the vertical shape changing means 35-1 and the sub-band division are output. Input to means 35-2. The vertical sub-band division has the same configuration and the same operation as the horizontal sub-band division, and performs sub-band division of an image signal having an arbitrary shape in 3n stages in the vertical direction. In this way, two-dimensional 3n-stage subband division can be performed by one memory storage process, and high-speed operation is possible.

  In addition, since the signal thus divided into sub-bands is divided into frequency bands as shown in FIG. 5, the low-frequency region is vertically and horizontally compared with the conventional sub-band division of FIG. Since the direction is also finely divided, the correlation within each frequency band is reduced, and the compression encoding efficiency is improved.

  <Effects> As described above, according to the fourth embodiment of the present invention, as in the third embodiment, only signals in an image area are sub-band-divided using shape information, and By changing the shape information to correspond to the image signal after the sub-band division, sub-band division at an arbitrary stage is possible for an image signal of an arbitrary shape. In addition, by changing the shape information, the subsequent subband division can always be performed on the same frequency component, and a subband division of an arbitrary shape can be performed without impairing the subband division efficiency. The subband division of n stages can be performed by one memory storage process, and the speed can be increased. In addition, since the low-frequency region of the sub-band divided signal is finely divided in the vertical and horizontal directions as shown in FIG. 5, the correlation in each frequency band is reduced, and the compression encoding efficiency is improved. .

<< Specific Example 5 >>
<Structure> In order to solve such a problem, a fifth embodiment of the present invention is configured as shown in FIG. Hereinafter, Embodiment 5 according to the present invention will be described in detail with reference to the drawings. The sub-band division method according to the fifth embodiment of the present invention includes a selector for selecting whether or not to perform horizontal (or vertical) sub-band division of the first stage and a sub-band division for performing one-dimensional sub-band division. A filter, a selector for selecting whether or not to perform vertical (or horizontal) sub-band division on the second stage, a sub-band division filter for performing one-dimensional sub-band division, a memory, and conversion control means ing.

  In FIG. 9, a two-dimensional image signal S8 sequentially and one-dimensionally input in the horizontal direction (or the vertical direction) and a two-dimensional image signal sequentially input in the horizontal direction (or the vertical direction) from the second-stage memory 76 are also input. The signal S76-1 of a certain frequency band of the image to be connected is connected to the first-stage selector 71, and the output S71-1 of the first-stage selector 71 performs the first-stage one-dimensional sub-band division. The output S71-2 of the first-stage selector 71 is directly connected to the first-stage memory 73, and the low-frequency component of the first-stage sub-band division filter is connected to the sub-band division filter 72. The output S72-1 and the output S72-2 of the high frequency component are connected to the first-stage memory 73, respectively, and are sequentially output one-dimensionally in the vertical (or horizontal) direction from the first-stage memory. Low frequency component S74 1 is connected to the second-stage selector 74-1. The high-frequency component S73-2 of the image sequentially and one-dimensionally output in the vertical (or horizontal) direction from the first-stage memory is The first outputs S74-1 and S74-3 of the second-stage selectors 74-1 and 74-2 are connected to the second-stage selector 74-2. 1 and 75-2, the second outputs S74-2 and S74-4 of the second-stage selectors 74-1 and 74-2 are connected to the second-stage memory 76, and The outputs S75-1 and S75-3 of the low frequency components and the outputs S75-2 and S75-4 of the high frequency components of the second subband splitting filters 75-1 and 75-2 are connected to the second-stage memory 76, respectively. And the output S7 of the second stage memory -1 is connected to the first stage selector 71, the output S76-2 of the memory of the second stage is connected to the output terminal. The conversion control signal S9 input from the outside is connected to the conversion control unit 77, the first output S77-1 of the conversion control unit 77 is connected to the first-stage selector 71, and the second output And third outputs S77-2 and S77-3 are connected to second-stage selectors 74-1 and 77-2, respectively.

  <Operation> A two-dimensional image signal S8 sequentially input one-dimensionally in the horizontal direction (or vertical direction) or a sub-band of an image sequentially output one-dimensionally in the horizontal direction (or vertical direction) from the second memory 76 When the signal S76-1 of a certain frequency band after the division is input, the first-stage selector 71 selects one of them according to a control signal S77-1 from a conversion control unit 77 described later, and It is output to the sub-band division filter 72 for performing one-dimensional sub-band division of the eye or directly to the memory 73 of the first stage. For example, in the first subband division, the image signal S8 from the outside is selected, and in the second and subsequent times, the signal S76-1 of a certain frequency band of the image from the memory 76 in the second stage is selected, and the horizontal direction is selected. When a (or vertical) one-dimensional wavelet is performed, the wavelet is output to the first-stage sub-band division filter 72, and otherwise, is directly output to the first-stage memory 73.

  The first-stage sub-band splitting filter 72 applies a one-dimensional horizontal (or vertical) direction to the image signal S71-1 input from the selector, as in Expressions 1 and 2, as in the related art. , And outputs the low frequency component S72-1 and the high frequency component S72-2 to the first stage memory 73.

  The first-stage memory 73 stores the input low-frequency component S72-1 and high-frequency component S72-2 in the horizontal direction (or vertical direction) or the image signal S71-2 input directly from the selector. The obtained image signals are transposed in the direction perpendicular to the input, and output one-dimensionally in the vertical direction (or the horizontal direction) to the second-stage selectors 74-1 and 74-2 in sequence.

  The second-stage selector 74-1 controls the low-frequency component S73-1 of the image that is output one-dimensionally in the vertical (or horizontal) direction from the memory in the first stage by a later-described conversion control unit 77. The signal is output to the sub-band division filter 75-1 for performing one-dimensional sub-band division of the second stage or the memory 76 of the second stage in accordance with the signal S77-2. For example, when a vertical (or horizontal) sub-band division is performed on an input signal, the signal is output to the second-stage sub-band division filter 75-1; 76.

  The second-stage selector 74-2 converts the high-frequency component S73-2 of the image output one-dimensionally in the vertical direction (or the horizontal direction) from the memory in the first stage in the same manner as 74-1. In accordance with the control signal S73-3 from the control means 77, the signal is output to the sub-band division filter 75-2 for performing the one-dimensional sub-band division of the second stage or the memory 76 of the second stage. For example, when a vertical (or horizontal) sub-band division is performed on an input signal, the signal is output to the second-stage sub-band division filter 75-2. Output.

  The second-stage sub-band division filters 75-1 and 75-2 perform one-dimensional sub-band division on the input image signal in the same manner as the first-stage sub-band division filter 72 to reduce the level. The frequency components S75-1 and S75-3 and the high frequency components S75-2 and S75-4 are output to the second-stage memory 76, respectively.

  In the memory 76 of the second stage, the image signals S75-1 and S75-2 and S75-3 and S75-4 sub-band divided by the sub-band division filters 75-1 and 75-2 of the second stage or The image signals S74-2 and S74-4 input directly from the second-stage selectors 74-1 and 74-2 are stored, and the signal S76-1 of each frequency band of the image is stored in the first-stage selector 71. Or, after all the conversions have been completed, output the sub-band divided image signal to an external terminal.

  The conversion control means 77 determines whether subband division should be performed on the image signal of interest or the frequency band signal of interest in the image in accordance with a control signal S9 input from the outside or a conversion method internally determined in advance. And outputs control signals S77-1 and S77-2 and S77-3 to the first-stage and second-stage selectors, respectively. For example, for a signal in a certain frequency band of an image to be divided into two-dimensional sub-bands, the signal is selected by the first-stage selector 71 and output to the first-stage sub-band division filter 72. Control signal S77-1 is output as described above. The outputs S73-1 and S73-2 from the first-stage memory 73 are supplied to the second-stage selectors 74-1 and 74-2, respectively. The control signals S77-2 and S77-3 are output so as to be output to 75-2. For an image signal to be subjected to horizontal (or vertical) one-dimensional sub-band division, the signal is selected by a first-stage selector 71 and output to a first-stage sub-band division filter 72. To do. Also, the outputs S73-1 and S73-2 from the first-stage memory 73 are directly output to the second-stage memory 76 for the second-stage selectors 74-1 and 74-2. I do. For an image signal to be divided in the vertical (or horizontal) one-dimensional sub-band, the first-stage selector 71 outputs the image signal directly to the first-stage memory 73, For the second-stage selectors 74-1 and 74-2, the outputs S73-1 and S73-2 from the first-stage memory 73 are respectively used as the second-stage sub-band division filters 75-1 and 75-2. Output to

  In this specific example, as shown in FIG. 23, the frequency band in which only the second-stage vertical (or horizontal) sub-band division is performed is directly read from the second-stage memory in the vertical (or horizontal) direction. ) Can be modified so that the frequency components S76-3 and S76-4 corresponding to ()) are called and output to the second-stage selectors 74-1 and 74-2.

  In this specific example, the first-stage memory and the second-stage memory can be shared by one memory. Further, since the first-stage selector and the second-stage selector have the same function, they can be shared by one selector. Furthermore, the first-stage sub-band splitting filter and the second-stage sub-band splitting filter have the same function, and can be shared by one sub-band splitting filter.

  <Effects> As described above, according to Embodiment 5 of the present invention, a two-dimensional image signal can be divided into an arbitrary band by inserting a selector and a conversion control unit into two-dimensional sub-band division. .

  For example, if the two-dimensional subband division is repeatedly performed on the low frequency components in both the vertical and horizontal directions, the image signal can be divided into the conventional frequency band of the image signal as shown in FIG. If the horizontal one-dimensional sub-band division is further performed on the low-frequency component only in the horizontal direction, the frequency band can be divided into frequency bands as shown in FIG. By further performing one-dimensional sub-band division in the vertical direction, the frequency band can be divided into frequency bands as shown in FIG. Further, if two-dimensional subband division is repeatedly performed on all frequency components, the image can be divided into image frequency bands as shown in FIG. The same frequency band division can be performed in the vertical direction.

  As described above, it is possible to divide an optimal frequency band according to the characteristics of an image, and to apply the method to image coding, the compression coding efficiency is improved.

<< Specific Example 6 >>
<Structure> Hereinafter, a sixth embodiment of the present invention will be described in detail with reference to the drawings. The image sub-band division method according to Embodiment 6 of the present invention is the same as the image sub-band division method according to Embodiment 5 of the present invention, except that the image is divided into a plurality of blocks in advance before frequency band division is performed. The subband division of the specific example 5 is performed independently.

  FIG. 13 is a configuration diagram of an image subband division method according to Embodiment 6 of the present invention. In FIG. 13, an input image signal is connected to a block dividing unit 10, and the blocks after the block dividing unit 10 are connected in the same manner as in the fifth embodiment.

  <Operation> In the image sub-band division method according to the sixth embodiment of the present invention shown in FIG. 13, when a two-dimensional image signal S1 is input, the block division means 10 converts the two-dimensional image into a plurality of N × M (N, The block is divided into M> 1), and each block signal S8 is output to the sub-band division method according to the fifth embodiment. The subband dividing means 81 of the specific example 5 independently and adaptively divides a frequency band for each block.

  <Effects> As described above, according to Embodiment 6 of the present invention, in Embodiment 5 of the present invention, by adding a block dividing unit, frequency band division can be performed independently in block units. For image blocks having high correlation in the vertical direction (for example, blocks having vertical edges), vertical sub-band division is performed, and image blocks having high correlation in the horizontal direction (for example, when horizontal edges are detected). For existing blocks), frequency band division can be performed according to local characteristics in the image by dividing the subband in the horizontal direction. Can be further enhanced.

<< Specific Example 7 >>
<Structure> Hereinafter, a seventh embodiment of the present invention will be described in detail with reference to the drawings. FIG. 14 is a configuration diagram of an image sub-band division method according to Embodiment 7 of the present invention. The image sub-band dividing method according to the seventh embodiment of the present invention includes two stages of image sub-band dividing means according to the fifth embodiment of the present invention and a block dividing means inserted therebetween.

  <Operation> In FIG. 14, the sub-band dividing means 91 for the image of the fifth embodiment of the present invention divides the frequency band of the entire image in the same manner as in the fifth embodiment. A considerable number of image signals are extracted from each frequency band of the frequency band-divided image, and are divided into a plurality of N × M blocks having the same band as the original frequency band divided image. For example, as shown in FIG. 15B, an image divided into bands as shown in FIG. 15A is extracted from each frequency band to form a corresponding image signal to form a block. The image sub-band dividing means 93 of the subsequent example 5 of the present invention performs adaptive frequency band division independently for each block.

  <Effects> As described above, according to the seventh embodiment of the present invention, the frequency band is divided according to the local characteristics in the image by dividing the block and performing frequency band division for each block independently. Division can be performed, and when applied to image coding, image coding efficiency can be further improved. Further, since the frequency band division is performed on the entire image before performing the block division, the reverse operation of the specific example 7 of the present invention is performed at the time of image decoding, so that the entire image is finally processed. A sub-band synthesis filter is performed, so that block noise due to encoding in block units can be eliminated, and a high-quality compressed and restored image can be obtained.

<< Specific Example 8 >>
<Construction> The image sub-band division method according to Embodiment 8 of the present invention is the same as the image sub-band division method according to Embodiments 5 to 7 described above, except that the second stage memory is used as shown in FIG. The configuration is obtained by adding a correlation calculating unit 18 that obtains vertical and horizontal correlations of a frequency band signal of interest of an input image and outputs the correlation values to a conversion control unit.

  <Operation> In FIG. 16, when a signal of a frequency band of interest of an image is input, the correlation calculating means 18 calculates the correlation between the input signals in the vertical and horizontal directions (for example, the similarity between adjacent image signals in the vertical and horizontal directions). ) And outputs the correlation value S18 to the conversion control means 77. When the correlation value (or its absolute value) is larger than a predetermined threshold, the conversion control means 77 further performs frequency band division in the direction (subband division in the direction). As in the specific example, the control signal S17-1 is output to the first-stage selector 71, and the control signals S17-2 and S17-3 are output to the second-stage selectors 74-1 and 14-2, respectively.

  <Effects> As described above, according to the eighth embodiment of the present invention, the division of the frequency band of an image can be determined by the correlation characteristics of each frequency band of the image. Are finely divided, and there is no redundancy. Further, since the frequency band is not forcibly divided at a place where the correlation is low, a spatial characteristic (for example, edge information or the like) is left, and coding efficiency can be improved.

<< Specific Example 9 >>
<Structure> The subband division method for an image according to the ninth embodiment of the present invention is the same as the subband division method for an image according to the sixth and seventh embodiments, as shown in FIG. A correlation calculating unit 18 for calculating the correlation in the vertical and horizontal directions of the band signal, and a memory 79 for storing the correlation value S18 and outputting the correlation value S79 of the neighboring block of the image block of interest to the conversion control unit 77 are added. It is configured.

  <Operation> In FIG. 17, when a certain frequency band signal S76-1 containing an image is input from the memory 76 at the second stage, the correlation calculating means 18 calculates the vertical and horizontal correlation of each frequency band of the block. Then, the correlation value S18 is stored in the memory 79. The conversion control means 77 calls the correlation value S79 of the corresponding frequency band of the block adjacent to the block of interest from the memory 79, and the vertical and horizontal correlation values (or their absolute values) are larger than a predetermined threshold value. In this case, the control signal S17-1 is supplied to the first-stage selector 71 in the same manner as in the embodiment 8 so that the frequency band is further divided in the corresponding direction of the target block (sub-band division in that direction). And outputs the control signals S17-2 and S17-3 to the second-stage selectors 74-1 and 14-2, respectively. For example, when the correlation in the vertical direction is large in the target frequency band of the block above the target block, the vertical frequency band division is further performed on the target frequency band of the target block. The same applies to the horizontal direction. Alternatively, when a large correlation exists in both directions, the frequency band is divided in both directions.

  <Effects> As described above, according to the ninth embodiment of the present invention, the division of the frequency band of the target block of the image can be determined by the correlation characteristics of each frequency band of the neighboring blocks of the image block. Where the frequency is high, the frequency band is finely divided and there is no redundancy. Further, since the frequency band is not forcibly divided at a place where the correlation is low, a spatial characteristic (for example, edge information or the like) is left, and coding efficiency can be improved. Further, since the encoding side determines the frequency band division number of the block of interest using the correlation value of the neighboring block of the block of interest, the decoding side similarly calculates the division number of the frequency band of the block of interest from the correlation value of the neighboring block. Since it can be obtained, it is not necessary to transmit the frequency band division information, and the code amount can be reduced.

<< Specific Example 10 >>
<Construction> In the tenth embodiment of the present invention, a filter coefficient having a rational number, a product of a rational number and a square root of 2 or an approximate value thereof is used in the subband division filter.

  <Operation> Each filter coefficient can be represented by Expression 3 or Expression 6 by making the filter coefficient a rational number or a product of a rational number and the square root of 2 or an approximate value thereof. In Equations 3 and 6, f (k) represents h (k) and g (k) in Equations 1 and 2. Ai (k) has a value of 0 or 1.

  Therefore, for example, if the filter coefficient is as shown in Expression 3, the operations of Expressions 1 and 2 can be performed by shift operation, addition, and one division, as shown in Expression 7, thereby increasing the speed. Is possible. Expression 7 shows that the operation of Wl + 1 (i) in Expression 2 is configured by shifting, adding, and one division, but the same applies to Sl + 1 (i). Further, in Equation 7, gs (k) has a value of 0 or 1, and therefore, it is sufficient to perform an operation only on a term having a value of 1.

Furthermore, if the coefficient b is a power of 2, the division in Equation 7 can be replaced with a shift operation, and all operations can be performed by shift operation and addition. Even when the filter coefficient is as shown in Expression 6, if the sub-band division (for example, two-dimensional sub-band division) is performed an even number of times, the multiplication of sqrt (2) is performed twice to obtain 2. Therefore, the shift operation can be performed. Therefore, subband division that does not require multiplication and division can be realized, and speeding up and simplification of the apparatus are possible. Examples of such a filter include the following. sqrt (x) indicates the square root of x.
h = (1, 1) sqrt (2) / 2g = (1, -1) sqrt (2) / 2h = (1, 1) sqrt (2) / 2g = (-1, -1, 8, -8 , 1, 1) sqrt (2) / 16h = (1, 1) sqrt (2) / 2g = (3, 3, -22, -22, 128, -128, 22, 22, -3, -3) sqrt (2) / 256h = (1, 2, 1) sqrt (2) / 4g = (-1, -2, 6, 2, 1) sqrt (2) / 8h = (1, 2, 1) sqrt ( 2) / 4g = (3, 6, -16, -38, 45, -45, 38, 16, -6, -3) sqrt (2) / 128h = (1, 2, 1) sqrt (2) / 4g = (-5, -10,34,78,123, -324,700, -324,123,78,34, -10, -5) sqrt (2) / 1024 These filters are all addition and shift operations Can perform the filter operation.

  <Effects> As described above, according to the tenth embodiment of the present invention, the sub-band division filter coefficients are rational numbers or the product of the rational number and the square root of 2 or those having an approximate value thereof. Shift operation, addition, and zero or one multiplication or division can be performed, so that high speed and simplification of the device can be realized. Therefore, the problem of the conventional technique can be solved.

<< Specific Example 11 >>
<Structure> Embodiment 11 of the present invention is a method of restoring an image with respect to the image sub-band division method of Embodiments 5, 6, and 8 of the present invention. Example 11 of the present invention is configured as shown in FIG. Hereinafter, Embodiment 11 of the present invention will be described in detail with reference to the drawings. The sub-band division restoration method according to the eleventh embodiment of the present invention includes a first-stage memory for storing sub-band divided images, and a first-stage horizontal (or vertical) sub-band synthesis filter. A selector group for selecting whether or not to perform the operation, a sub-band synthesis filter group for performing a one-dimensional sub-band synthesis filter, a second-stage memory for storing the results of the first-stage sub-band synthesis filter, and a second-stage memory A selector for selecting whether or not to perform a vertical (or horizontal) sub-band synthesis filter on the eyes, a sub-band synthesis filter for performing a one-dimensional one-dimensional sub-band synthesis filter on the second stage, and a second-stage weblet A third stage memory for storing the result of the inverse conversion and an inverse conversion control unit are provided.

  In FIG. 18, a two-dimensional image S40 divided from sub-bands input from the outside or an image signal S47-1 output from a memory 47 in a third stage described later is connected to a memory 41 in the first stage. A low-frequency component or a component S41-1 at the same position in both the vertical and horizontal directions of the two-dimensional image signal sequentially and horizontally output from the memory 41 of the first stage in the horizontal direction (or vertical direction) and the horizontal direction (or vertical direction) Is a low-frequency component or a component S41-2 at the same position is connected to the first-stage selector 42-1 and sequentially output one-dimensionally in the horizontal (or vertical) direction from the first-stage memory 41. In the two-dimensional image signal, a low-frequency component or a component S41-3 at the same position in the vertical direction (or the horizontal direction) and a high-frequency component or a component S41-4 at the same position in the horizontal and vertical directions are provided in the first stage. The first output S42-1 and the second output S42-2 of the first-stage selector 42-1 are connected to the first-stage one-dimensional sub-band synthesis filter. The third output S42-3 of the first-stage selector 42-1 is connected to the second-stage memory 44, and the third output S42-3 of the first-stage selector 42-1 is connected to the second-stage memory 44. The first output S43-1 is connected to the second-stage memory 44, and the first output S42-4 and the second output S42-5 of the first-stage selector 42-2 are connected to the first-stage memory 42. , And a third output S42-6 of the first-stage selector 42-2 is connected to the second-stage memory 44, Output S of first-stage subband synthesis filter 43-2 3-2 are connected to the second-stage memory 44, and the low-frequency component S44-1 and the high-frequency component of the image sequentially output one-dimensionally in the vertical (or horizontal) direction from the second-stage memory 44 S44-2 is connected to the second-stage selector 45, and the first outputs S45-1 and S45-2 of the second-stage selector 45 are connected to the second-stage subband synthesis filter 46. The third output S45-3 of the second-stage selector 45 is connected to the third-stage memory 47, and the output S46 of the second-stage sub-band synthesis filter 46 is connected to the third-stage memory. 47, the output S47-1 of the third-stage memory is connected to the first-stage memory 41, and the output S76-2 of the third-stage memory is connected to the output terminal. The conversion control signal S3 input from the outside is connected to the inverse conversion control means 48, and the first and second outputs S48-1 and S48-2 of the conversion control means 48 are connected to the first-stage selector 42-1 and The second output S48-3 of the conversion control means 48 is connected to the second-stage selector 45.

  <Operation> When a two-dimensional image signal S40 divided into subbands from the outside or a partially restored two-dimensional image signal output from the memory 47 in the third stage is input, the first stage In the memory 41, the two-dimensional image is temporarily stored, and a signal S41-1 (for example, divided into frequency bands as shown in FIG. 5) in an upper left quarter of a frequency band to be subjected to inverse subband division When the image is input, the signal LL1 of the low frequency component in both the vertical and horizontal directions is sequentially output to the first-stage selector 42-1 in the horizontal direction (or the vertical direction) one-dimensionally. The signal S41-2 (for example, the signal of the low frequency component HL1 in the horizontal direction in FIG. 5) of the one-half area is sequentially and one-dimensionally output to the first-stage selector 42-1 in the horizontal direction (or the vertical direction). , The signal S41-3 in the lower left quarter region ( For example, the signal of the low frequency component LH1 in the vertical direction in FIG. 5 is sequentially output in the horizontal direction (or the vertical direction) one-dimensionally to the first-stage selector 42-2, and the lower right quarter region is output. The signal S41-4 (for example, a signal of the high-frequency component HH1 in both the vertical and horizontal directions in FIG. 5) is sequentially and one-dimensionally output to the first-stage selector 42-2 in the horizontal direction (or the vertical direction).

  The first-stage selector 42-1 converts the image signals S41-1 and S41-2 inputted from the first-stage memory in accordance with the inverse transformation control signal S48-1 inputted from the inverse transformation control means 48, respectively. The data is output to the subband synthesis filter 43-1 as S42-1 and S42-2, or is output to the second-stage memory 44 as S42-3 in the order of S41-1 and S41-2.

  In the first-stage selector 42-2, similarly to 42-1, the image signal S41- inputted from the first-stage memory according to the inverse transformation control signal S48-2 inputted from the inverse transformation control means 48. 3 and S41-4 are output to the subband synthesis filter 43-2 as S42-4 and S42-5, respectively, or are output to the second-stage memory 44 as S42-6 in the order of S41-3 and S41-4. I do.

  The first-stage subband synthesis filters 43-1 and 43-2 perform inverse subband division on the input low-frequency component S42-1 or S42-4 and the high-frequency component S42-2 or S42-5. , And outputs the resulting image signal S43-1 or S43-2 to the second-stage memory 44. The sub-band synthesis filter includes a pair of a low-pass filter 35-1, a high-pass filter 35-2, and an adder 36 as shown in FIG. When the low-frequency component S34-1 is input, the low-pass filter first inserts one 0 for every other signal into the input signal as shown in Expression 8, and applies a low-pass filter as shown in Expression 10. . Similarly, in the high-pass filter, one 0 is inserted for every other signal in the input high-frequency component as in Expression 9, and the high-pass filter is applied as in Expression 11. In Expressions 8 to 11, S (i) and W (i) represent a low-frequency component and a high-frequency component, respectively, and h ′ (k) and g ′ (k) represent coefficients of a low-pass filter and a high-pass filter, respectively.

  The adder 36 adds the result S ″ (i) of the low-pass filter and the result W ″ (i) of the high-pass filter to complete a one-dimensional sub-band synthesis filter and restore the image signal. The restored signal becomes the low frequency component of the next wavelet layer. Therefore, the horizontal (or vertical) low frequency components are restored by the first-stage subband synthesis filter.

  In the second-stage memory 44, the low-frequency component S43-1 and the high-frequency component S43-2 in the vertical direction (or the vertical direction) input from the sub-band synthesis filter or the equivalent input from the first-stage selector are input. The image signals S42-3 and S42-6 are stored, and the input image signals are transposed in a direction perpendicular to the input, and are sequentially and one-dimensionally vertically (or horizontally) shifted to the second stage selector-45. Output to

  In the second-stage selector 45, as in the first-stage selectors 42-1 and 42-2, the second-stage memory is operated in accordance with the inverse conversion control signal S48-3 input from the inverse conversion control means 48. The input image signals S44-1 and S44-2 are output to the sub-band synthesis filter 46 as S45-1 and S45-2, respectively, or in the order of S44-1 and S44-2 as S45-3 in the third stage. Output to the memory 47 of the eye.

  In the second-stage sub-band synthesis filter 46, the input low-frequency component S45-1 and high-frequency component S45-2 are processed in the same manner as the first-stage sub-band synthesis filters 43-1 and 43-2. The inverse subband division is performed, and the resulting image signal S46 is output to the memory 47 in the third stage. As described above, the low-frequency component in the vertical direction (or the horizontal direction) is restored by the second-stage sub-band synthesis filter, so that the first-stage horizontal (or vertical) sub-band synthesis filter is used. Together, two-dimensional inverse subband division is realized. For example, when a one-stage two-dimensional sub-band synthesis filter is applied to the image signal divided into the frequency bands in FIG. 5, a part of the frequency band is restored, and the frequency band is divided as shown in FIG.

  The third-stage memory 47 temporarily stores the output S46 of the second-stage subband synthesis filter 46 or the image signal S45-3 input from the second-stage selector 45, and After the sub-band synthesis filter is completed, the image signal S47-1 is output to the first-stage memory 41 again, or after all conversions are completed, the image signal S47-2 is output to an external terminal.

  The inverse transformation control means 48 determines whether or not to perform the inverse sub-band division on the image signal of interest or the frequency band signal of interest of the image according to the control signal S3 input from the outside, and the first-stage selector Control signals S48-1 and S48-2 are output to 42-1 and 42-2, and a control signal S48-3 is output to the second-stage selector 45.

  The above operation is repeated until all the image signals are restored.

  In this specific example, as shown in FIG. 24, for the frequency band in which only the vertical (or horizontal) sub-band synthesis filter of the second stage is used, the vertical (or horizontal) The direction can be modified such that the frequency component S47-3 corresponding to (direction) is called and output to the second-stage memory 44.

  In this specific example, the first-stage memory, the second-stage memory, and the third-stage memory can be shared by one memory. Further, since the first-stage selector and the second-stage selector have the same function, they can be shared by one selector. Further, the first-stage sub-band synthesis filter and the second-stage sub-band synthesis filter also have the same function, and can be shared by one sub-band synthesis filter.

  <Effects> As described above, according to Embodiment 11 of the present invention, a two-dimensional image signal divided into an arbitrary band can be obtained by inserting a selector and an inverse transformation control unit into two-dimensional inverse sub-band division. Can be restored. The method for restoring a subband divided image according to the eleventh embodiment of the present invention can restore the subband divided images according to the fifth, sixth, and eighth specific examples.

<< Specific Example 12 >>
<Structure> Embodiment 12 of the present invention is a method of restoring an image to the image sub-band division method of Embodiment 7 of the present invention. Embodiment 12 of the present invention is configured as shown in FIG. The details will be described below with reference to the drawings. The method for restoring a sub-band divided image according to the twelfth embodiment of the present invention is configured by inserting two stages of restoring means for a sub-band divided image according to the eleventh embodiment of the present invention and inserting an image reconstructing means therebetween.

  <Operation> In FIG. 19, in the sub-band divided image restoration means 61 in the preceding stage, the image is reconstructed as a block by the above-described specific example 7 and the frequency band is divided into the respective blocks. The image is independently restored for each block, and the image reconstructing means 63 collects the same frequency components in each block for the image partially restored independently for each block, and Reconfigure the band. For example, an image in which blocks are reconstructed as shown in FIG. 15B is reconstructed as shown in FIG. Subsequent sub-band divided image restoration means 62 restores the reconstructed frequency band divided image in the same manner as in Example 11.

  <Effects> As described above, according to the twelfth embodiment of the present invention, the image which is reconstructed by the block according to the seventh embodiment of the present invention and divided into frequency bands independently for each block is firstly processed. By performing adaptive image restoration on a block-by-block basis, reconstructing an image, and performing image restoration on the whole, it is possible to eliminate block noise caused by block-by-block coding while maintaining high coding efficiency. A high quality restored image can be obtained.

<< Specific Example 13 >>
<Structure> Embodiment 13 of the present invention is a method of restoring an image with respect to the image subband dividing method of Embodiment 10 of the present invention. Example 13 of the present invention is configured as shown in FIG. The method for restoring a sub-band split image according to the thirteenth embodiment of the present invention is the same as the method for restoring a sub-band split image according to the eleventh embodiment, as shown in FIG. It is configured by adding a correlation calculating means 49 for obtaining a correlation and a memory 50 for storing the correlation value S49 and outputting a correlation value S50 of a neighboring block of the image block of interest to the conversion control means 48. .

  <Operation> In FIG. 20, when a certain frequency band signal S47-1 including a block image is input from the memory 47 at the third stage, the correlation calculating means 49 performs the same operation as the correlation calculating means of the specific example 10 described above. The correlation in the vertical and horizontal directions of each frequency band of the block is obtained, and the correlation value S49 is stored in the memory 50. The conversion control means 48 calls the correlation value S50 of the corresponding frequency band of the block adjacent to the block of interest from the memory 50, and the vertical and horizontal correlation values (or their absolute values) are larger than a predetermined threshold value. In this case, it is determined that the pertinent direction of the block of interest is divided into frequency bands (subband division is performed in that direction), and the control signals S48-1 and S48-2 are used to perform image restoration in the pertinent direction. The control signal S48-3 is output to the first-stage selectors 42-1 and 42-2, and the control signal S48-3 is output to the second-stage selector 45. For example, when the vertical correlation is large in the target frequency band of the block above the target block, the vertical image restoration is performed on the target frequency band of the target block. The same applies to the horizontal direction. Alternatively, when a large correlation exists in both directions, image restoration is performed in both directions.

  <Effects> As described above, according to the thirteenth embodiment of the present invention, the image restoration of the target block of the image can be determined by the correlation characteristics of each frequency band of the neighboring blocks of the image block. Even without the above, it is possible to adaptively restore an image that has been frequency-band divided, and to reduce the amount of code.

<< Specific Example 14 >>
<Configuration> In Embodiment 14 of the present invention, a filter coefficient having a rational number, a product of a rational number and a square root of 2 or an approximate value thereof is used in a subband synthesis filter.

  <Operation> Each filter coefficient can be represented by Expression 3 or Expression 6 by making the filter coefficient a rational number or a product of a rational number and the square root of 2 or an approximate value thereof. Here, f (k) in Expressions 3 and 6 represents h ′ (k) and g ′ (k) in Expressions 10 and 11. Ai (k) has a value of 0 or 1. Similarly to the specific example 10, if, for example, the filter coefficients are as shown in Expression 3, the operation of the inverse subband division of Expressions 10 and 11 is performed as shown in Expression 12 by shift operation, addition, and 1 It can be performed by dividing the number of times, and the speed can be increased. Equation 12 shows that the operation of W ″ l (i) in equation 11 is configured by shifting, adding, and one division, but the same applies to S ″ l (i). In Equation 12, g's (k) has a value of 0 or 1 and is 0 every other W'l (i) and S'l (i). The operation need only be performed on the term having.

  Furthermore, if the coefficient b is a power of 2, the division in Equation 12 can be replaced with a shift operation, and all operations can be performed by shift operation and addition. Even when the filter coefficient is as shown in Expression 6, if the number of subband synthesis filters is limited to an even number (for example, a two-dimensional subband synthesis filter), multiplication of sqrt (2) is performed twice. Since it is 2, the shift operation can be performed. Therefore, a sub-band synthesis filter that does not require multiplication and division can be realized, and speeding up and simplification of the device can be achieved. As such a filter, for example, the filter of Example 10 can be used to obtain the following.

  <Effects> As described above, according to the fourteenth embodiment of the present invention, the sub-band synthesis filter coefficient is obtained by using a rational number, a product of a rational number and a square root of 2 or an approximate value thereof. Can be performed by shift operation, addition, and zero or one multiplication or division, thereby realizing high-speed operation and simplification of the device. Therefore, the problem of the conventional technique can be solved.

FIG. 3 is a configuration diagram of a subband dividing device according to a first embodiment of the present invention. FIG. 2 is a configuration diagram of a conventional subband splitting device. FIG. 3 is a configuration diagram of a subband division filter. It is a frequency band division | segmentation figure after the subband division | segmentation of a prior art. It is a frequency band division figure after subband division of the present invention. FIG. 9 is a configuration diagram of a subband splitting device of Embodiment 2 according to the present invention. FIG. 13 is a configuration diagram of a shape adaptive subband dividing apparatus according to a third embodiment of the present invention. FIG. 13 is a configuration diagram of a shape adaptive sub-band division device according to a fourth embodiment of the present invention. FIG. 14 is a configuration diagram of a subband division method according to Embodiment 5 of the present invention. 5 is a first example of frequency band division by the sub-band division method of the present invention. 5 is a second example of frequency band division by the sub-band division method of the present invention. 9 is a third example of frequency band division by the sub-band division method of the present invention. FIG. 14 is a configuration diagram of a subband division method according to Embodiment 6 of the present invention. FIG. 16 is a configuration diagram of a subband division method of Embodiment 7 according to the present invention. FIG. 24 is an explanatory diagram of a block reconfiguration according to a seventh embodiment of the present invention. FIG. 19 is a configuration diagram of a subband division method of Embodiment 8 according to the present invention. FIG. 19 is a configuration diagram of a subband division method according to a ninth embodiment of the present invention. FIG. 21 is a configuration diagram of a method of restoring a subband divided image according to Embodiment 11 of the present invention. FIG. 21 is a configuration diagram of a method of restoring a subband divided image according to Example 12 of the present invention. FIG. 21 is a configuration diagram of a method for restoring a subband divided image according to Example 13 of the present invention. FIG. 3 is a configuration diagram of a subband synthesis filter according to the present invention. FIG. 6 is a configuration diagram after one-stage subband synthesis according to the present invention. FIG. 15 is a configuration diagram of a modification of the fifth embodiment according to the present invention. FIG. 21 is a configuration diagram of a modification of the eleventh embodiment according to the present invention.

Explanation of reference numerals

11-13, 15-17, 20, 22-1, 22-2, 24, 26-1, 26-2, 28, 30-1, 30-2 Sub-band division filters 14, 21-1, 21-2 , 23,25-1,25-2,27,29-1,29-2 memory

Claims (8)

  A sub-band division filter composed of a rational number or a product of a rational number and a square root of 2 or a filter coefficient having an approximate value thereof, wherein a sub-band division operation is performed by using addition and subtraction and a shift operation and one multiplication or division. Sub-band splitting filter.   The subband division filter according to claim 1, wherein the rational number is a rational number having a denominator of a power of 2, and the subband division operation is performed using only addition, subtraction, and shift operations.   A subband synthesis filter comprising a rational number or a product of a rational number and the square root of 2 or a filter coefficient having an approximation thereof, wherein a subband synthesis operation uses addition and subtraction and a shift operation and one multiplication or division. A sub-band synthesis filter.   The subband synthesis filter according to claim 3, wherein the rational number is a rational number having a denominator of a power of 2, and the subband synthesis operation is performed using only addition, subtraction, and shift operations.   The sub-band division filter includes operation means including a shift operation unit for performing an operation, a multiplier, a divider, and an adder. The operation means includes a rational number or a rational number of 2 as a sub-band division filter coefficient. A sub-band division filter that performs sub-band division by shift operation and addition and zero or one multiplication or division by using a product of square roots or an approximation thereof.   The sub-band division filter includes operation means including a shift operation unit for performing an operation and an adder. The operation means includes a rational number having a power of 2 in a denominator or a 2 in a denominator as a sub-band division filter coefficient. A subband division filter that performs a subband division by a shift operation and an addition by using a product of a rational number having a power and a square root of 2 or an approximate value thereof.   In the sub-band synthesis filter, the sub-band synthesizing filter includes operation means including a shift operation unit for performing an operation, a multiplier, a divider, and an adder, and the operation means includes a rational number or a rational number of 2 A subband synthesis filter characterized in that subband synthesis is performed by a shift operation and addition and zero or one multiplication or division by using a product of square roots or an approximation thereof.   The subband synthesizing filter includes a calculating unit including a shift calculating unit for performing a calculation and an adder, and the calculating unit includes a rational number having a power of 2 in a denominator or a 2 in a denominator as a subband dividing filter coefficient. A subband synthesis filter that performs a subband synthesis by a shift operation and an addition by using a product of a rational number having a power and a square root of 2 or an approximate value thereof.

Images