JP4371070B2 - Image layer encoding method and image layer decoding method - Google Patents

Description

  The present invention relates to an image layer encoding method and an image layer decoding method, and more particularly to an image for efficiently utilizing the capacity of communication and recording media, characterized in that the resolution can be easily changed by adding and deleting encoding units. The present invention relates to a hierarchical encoding method and an image hierarchical decoding method.

  In recent years, for the purpose of saving recording capacity and effective use of a transmission band, a high-efficiency encoding method using redundancy and statistical bias of image signals has become an indispensable technique. For example, MPEG2 (Moving Picture Experts Group 2) is widely used in moving picture digital recording devices and digital broadcasting. For still images, the JPEG 2000 (Joint Photographic Experts Group 2000) method, which improves the geometric distortion described later, is used.

  The MPEG2 system is an irreversible compression encoding system characterized by dividing an image into blocks and performing uneven distribution and quantization of data by discrete cosine transform. For this reason, when the compression rate is increased, geometric distortion that is easily recognized by humans, called block distortion and mosquito noise, is generated, leading to degradation of image quality. In view of these problems, the JPEG2000 system uses a discrete wavelet transform (hereinafter referred to as DWT) for orthogonal transform, and employs a hierarchical coding system based on subband division (see, for example, Non-Patent Document 1). ). Hereinafter, a subband hierarchical encoding method using DWT employed in the JPEG2000 method according to the present invention will be described.

  FIG. 21 is a block diagram showing the basic configuration of a DWT encoding apparatus. In the figure, an input image signal comes from an input terminal 100, undergoes signal format conversion such as image color conversion by an image input processor 101, and is then input to a DWT unit 102. In the DWT unit 102, the image signal output from the image input processor 101 is hierarchically divided into subbands and input to the quantizer 103.

  The quantizer 103 quantizes the input subband signal and inputs the quantized signal to the entropy encoder 104. The entropy encoder 104 encodes the quantized subband signal using a Huffman code or an arithmetic code to obtain encoded data. Finally, the encoded data output from the entropy encoder 104 is converted into a format for transmission or recording by the code output processor 105 and output to the output terminal 106.

  Next, a decoding apparatus for decoding a data string transmitted or recorded by the encoding apparatus in FIG. 21 will be described with reference to FIG. In FIG. 22, the transmitted or reproduced data sequence is input from the input terminal 200, and the entropy encoded data is extracted from the data sequence by the code input processor 201. The data output from the code input processor 201 is decoded into a quantized DWT coefficient sequence by the entropy decoder 202 of FIG. 22 that performs a complementary operation with the entropy encoder 104 of FIG.

  The quantized DWT coefficient sequence is dequantized by the inverse quantizer 203 of FIG. 22 that performs complementary operation with the quantizer 103 of FIG. Data). The hierarchized DWT coefficient sequence output from the inverse quantizer 203 is supplied to the IDWT unit 204, subjected to inverse discrete wavelet transform (hereinafter referred to as IDWT), decoded into an image signal, and then output as an image. The processor 205 performs signal format conversion such as image color conversion and outputs the result from the output terminal 206.

  As described above, subband hierarchical encoding and decoding are performed. Hereinafter, subbanding and hierarchization by the DWT unit 102 of FIG. 21, which is the main part of the present invention, will be described in detail. Since subbanding by DWT is divided into octaves, the image size is reduced horizontally and vertically in units of powers of 2. Octave decomposition is realized by an operation of applying a low-pass DWT filter (LPF) and a high-pass DWT filter (HPF) in the horizontal (H) / vertical (V) direction of a two-dimensional image, and downsampling.

  FIG. 23 shows a block diagram of an example of a vertical decomposer 300 that performs a vertical operation. In FIG. 23, a signal coming from the input terminal 301 is input to the divided V-LPF unit 302 and the divided V-HPF unit 303. The divided V-LPF unit 302 filters the vertical low frequency of the input signal and outputs it. Further, the divided V-HPF unit 303 filters and outputs the vertical high frequency of the input signal. The low frequency is a frequency on the lower side of the center frequency of the octave-decomposed band, and the high frequency is a frequency on the higher side of the center frequency.

  The vertical low-frequency signal output from the divided V-LPF unit 302 is input to the down-sampling unit 304, and the vertical high-frequency signal output from the divided V-HPF unit 303 is input to the down-sampling unit 305. The operation of thinning out every other sample is performed. As a result, a vertical low frequency signal (L) in which the number of samples in the vertical direction is halved by the downsampler 304 is output from the output terminal 306. Further, the vertical high frequency signal (H) in which the number of samples in the vertical direction is halved by the downsampling device 305 is output from the output terminal 307.

  FIG. 24 shows a block diagram of an example of a horizontal decomposer 400 that performs a horizontal operation. In the figure, a signal coming from an input terminal 401 is input to a divided H-LPF unit 402 and a divided H-HPF unit 403. The divided H-LPF unit 402 filters and outputs the horizontal low frequency of the input signal. The split H-HPF unit 403 filters and outputs the horizontal high frequency of the input signal.

  The horizontal low-frequency signal output from the divided H-LPF unit 402 is input to the down-sampling unit 404, and the horizontal high-frequency signal output from the divided H-HPF unit 403 is input to the down-sampling unit 405, respectively. The operation of thinning out every other sample is performed. As a result, a horizontal low frequency signal (L) in which the number of samples in the horizontal direction is halved by the downsampling unit 404 is output from the output terminal 406. Further, a horizontal high-frequency signal (H) in which the number of samples in the horizontal direction is halved by the down-sampling device 405 is output from the output terminal 407.

  Next, the basic unit of the restoration operation in the IDWT unit 204 in FIG. 22 for restoring the subband signal will be described with reference to FIGS. 25 and 26. FIG. In other words, the IDWT unit 204 performs a restoration operation based on the basic unit of the vertical synthesis by the vertical synthesizer 500 shown in FIG. 25 and the horizontal synthesis by the horizontal synthesizer 600 shown in FIG.

  FIG. 25 shows a block diagram of an example of a vertical synthesizer 500. The octave-resolved vertical low-frequency signal (L) enters from the input terminal 501 and is input to the upsampler 503. Further, the vertical high frequency signal (H) subjected to octave decomposition enters from the input terminal 502 and is input to the upsampler 504. Up-samplers 503 and 504 interpolate zero values every other sample in the vertical direction of the input signal. As a result, the upsamplers 503 and 504 output signals in which the number of samples in the vertical direction is doubled.

  The signal output from the upsampler 503 is input to the synthesis V-LPF unit 505, and the vertical low frequency is filtered. Further, the signal output from the upsampling unit 504 is input to the synthesis V-HPF unit 506 and the vertical high frequency is filtered. The vertical low frequency signal output from the combined V-LPF 505 and the vertical high frequency signal output from the combined V-HPF unit 506 are input to the adder 507 and added together. As a result, a signal having a restored vertical component is extracted from the adder 507 and output from the output terminal 508.

  FIG. 26 shows a block diagram of an example of a horizontal synthesizer 600. The horizontal low frequency signal (L) subjected to the octave decomposition is input from the input terminal 601 and input to the upsampler 603. Further, the horizontal high frequency signal (H) subjected to the octave decomposition comes from the input terminal 602 and is inputted to the upsampler 604. Up-samplers 603 and 604 interpolate zero values every other sample in the horizontal direction of the input signal. As a result, the upsamplers 603 and 604 output a signal in which the number of samples in the horizontal direction is doubled.

  The signal output from the upsampler 603 is input to the synthesis H-LPF unit 605 and the horizontal low frequency is filtered. In addition, the signal output from the upsampling device 604 is input to the combined H-HPF device 606 and the horizontal high frequency is filtered. The horizontal low frequency signal output from the combined H-LPF unit 605 and the horizontal high frequency signal output from the combined H-HPF unit 606 are input to the adder 607 and added. As a result, a signal having the restored horizontal component is extracted from the adder 607 and output from the output terminal 608.

  By the way, in the JPEG2000 system, as shown in the horizontal / vertical decomposer 700 in FIG. 27, the vertical vertical decomposer 300 in FIG. 23 and the horizontal horizontal decomposer 400 in FIG. The octave decomposition of is realized. In the following description, the same number is assigned to blocks having the same function.

  The image signal that has entered from the input terminal 701 in FIG. 27 is input to the input terminal 401 of the horizontal decomposer 400 and is subjected to octave decomposition in the horizontal direction. Of the signals subjected to octave decomposition by the horizontal resolver 400, the horizontal low frequency signal is output from the output terminal 406, and the horizontal high frequency signal is output from the output terminal 407. The horizontal low-frequency signal output from the output terminal 406 is input to the input terminal 301-1 of the vertical decomposer 300-1 and is subjected to octave decomposition in the vertical direction. Of the signals subjected to octave decomposition by the vertical decomposer 300-1, the vertical low frequency signal is output from the output terminal 306-1 and the vertical high frequency signal is output from the output terminal 307-1.

  As a result, the signal output from the output terminal 306-1 is output to the outside from the output terminal 702 of the horizontal / vertical decomposer 700 as a horizontal low-frequency / vertical low-frequency signal (LL). The signal output from the output terminal 307-1 is output to the outside from the output terminal 703 as a horizontal low frequency / vertical high frequency signal (LH).

  Similarly, the horizontal high-frequency signal output from the output terminal 407 is input to the input terminal 301-2 of the vertical decomposer 300-2 and is subjected to octave decomposition in the vertical direction. As a result, the signal output from the output terminal 306-2 is output to the outside from the output terminal 704 of the horizontal / vertical decomposer 700 as a horizontal high frequency / vertical low frequency signal (HL). The signal output from the output terminal 307-2 is output from the output terminal 705 to the outside as a horizontal high frequency / vertical high frequency signal (HH). In the following description, the above-described components LL, HL, LH, and HH are described by adding a hierarchical number to which each component belongs.

  FIG. 31 is a diagram schematically showing each component (LL1, LH1, HL1, HH1) of the first layer divided into four by the horizontal / vertical decomposer 700 of FIG. Among these, the second layer is created by further octave-decomposing the LL1 component which is a horizontal low-frequency / vertical low-frequency signal. FIG. 32 is a diagram schematically showing a state in which the LL1 component is further divided into four and created up to the second layer (LL2, LH2, HL2, HH2). Furthermore, it is possible to increase the number of layers such as the third layer by octave decomposition of the LL2 component.

  FIG. 28 shows means for dividing up to the second hierarchy in FIG. The image signal input to the input terminal 701-1 of the first horizontal / vertical decomposer 700-1 in FIG. 28 is converted into the first signal shown in FIG. 31 by the horizontal / vertical decomposer 700-1 by the operation described in FIG. The components LL1, LH1, HL1, and HH1 divided into four layers are output from output terminals 702-1, 703-1, 704-1, and 705-1, respectively. Among these, the LL1 component output from the output terminal 702-1 is input to the input terminal 701-2 of the second horizontal / vertical decomposer 700-2, and is divided into four components in the second hierarchy shown in FIG. LL2, LH2, HL2, and HH2 are output from output terminals 702-2, 703-2, 704-2, and 705-2, respectively. In this way, hierarchization is realized by adopting a configuration in which the LL component output from the horizontal / vertical decomposer is input to the next horizontal / vertical decomposer.

  In the JPEG2000 system, as shown in the horizontal / vertical synthesizer 900 in FIG. 29, the vertical synthesizer 500 in FIG. 25 described above and the horizontal synthesizer 600 in the horizontal direction in FIG. 26 are combined. , The octave decomposed signal is restored. In the horizontal / vertical synthesizer 900 shown in FIG. 29, the components LL, LH, HL, and HH subjected to octave decomposition come from the corresponding input terminals 901, 902, 903, and 904, respectively.

  An input terminal 501-1 of the vertical synthesizer 500-1 receives an LL component, an input terminal 502-1 receives an LH component, and a horizontal low-frequency signal obtained by combining the vertical components is output from an output terminal 508-1. Is output. Further, the HL component is input to the input terminal 501-2 of the vertical synthesizer 500-2, the HH component is input to the input terminal 502-2, and a horizontal high-frequency signal obtained by combining the vertical components is output from the output terminal 508-2. Is output.

  The horizontal low frequency signal output from the output terminal 508-1 is input to the input terminal 601 of the horizontal synthesizer 600, and the horizontal high frequency signal output from the output terminal 508-2 is input to the input terminal 602 of the horizontal synthesizer 600. Entered. The horizontal synthesizer 600 outputs a signal obtained by synthesizing the horizontal low-frequency signal and the horizontal high-frequency signal from the output terminal 608 and outputs it as an output signal of the horizontal / vertical synthesizer 900 from the output terminal 905 to the outside.

  Here, when restoring the hierarchized signal as shown in FIG. 32, the signal restored by the first horizontal / vertical synthesizer 900-1 is converted into the second horizontal / vertical signal as shown in FIG. This is realized by inputting to the input terminal 901-2 of the LL component of the combiner 900-2.

  FIG. 30 shows a block diagram of an example of a circuit for restoring the hierarchized signal of FIG. In FIG. 30, LL2, LH2, HL2, and HH2 components are input to input terminals 901-1, 902-1, 903-1, and 904-1 of the first horizontal / vertical combiner 900-1. The first horizontal / vertical synthesizer 900-1 restores the LL1 component based on the input components LL2, LH2, HL2, and HH2 by the operation described with reference to FIG. 29, and outputs it from the output terminal 905-1.

  The LL1 component output from the output terminal 905-1 is input to the input terminal 901-2 of the second horizontal / vertical synthesizer 900-2 and simultaneously input terminals 902-2, 903-2, and 904-2. Are input with LH1, HL1, and HH1 components. The second horizontal / vertical synthesizer 900-2 synthesizes the input components by the operation described with reference to FIG. 29, and outputs the restored image from the output terminal 905-2. Similarly, even in the case of a signal that is octave-decomposed into three or more layers, an image can be restored by cascade-connecting the horizontal / vertical synthesizer 900.

  At this time, the LL1 component output from the output terminal 905-1 of the first horizontal / vertical synthesizer 900-1 in FIG. 30 is halved in the number of pixels in the horizontal / vertical direction with respect to the original image size. It has become a reduced image. Therefore, the present hierarchical coding has a feature that a reduced image in units of powers of 2 can be obtained by presenting the output signal of the horizontal / vertical synthesizer 900 of each hierarchy. On the other hand, there is also a problem that only a reduced image in units of powers of 2 can be obtained.

  For this reason, a method for creating a reduced image of an arbitrary size has been proposed (see, for example, Patent Document 1 and Patent Document 2). Patent Document 1 discloses an image hierarchical decoding method for obtaining an image with an arbitrary resolution by extracting and adding a high-frequency component to be added to an LL component only by a necessary band by filtering. Further, Patent Document 2 discloses an image hierarchy decoding method in which a component that can obtain an image that is one layer larger than a required image size is restored, and a low-pass filter is applied to obtain the required image size.

“Next Generation Image Coding System JPEG2000”, Trikes, Inc. (ISDB4-88657-209-X) JP 2000-125294 A JP 2004-147095 A

  By the way, when transmitting or recording the encoded data, it is required to adjust the code amount so as to be within the transmission capacity or the recording medium capacity. As a means for adjusting the code amount, a method of coarsening the quantization width or a method of transmitting or recording only a part of the code subjected to octave decomposition can be considered. Among these methods, the method of transmitting or recording only a part of the code subjected to octave decomposition reduces the code in units of powers of 2, so that fine adjustment is not effective compared to the method of coarsening the quantization width. There is a problem.

  For example, when the code amount is slightly over the required transmission capacity or recording capacity, the method of reducing the code subjected to the octave decomposition may be used once even though the capacity may be reduced slightly. Therefore, a large amount of code is reduced, and the resolution is drastically reduced. For this reason, it is common to prioritize the code amount adjustment based on the quantization width, despite the advantage that the degradation of the visual characteristics is less noticeable than the method of reducing the quantization width. It is the target.

  Therefore, if the code amount can be adjusted more finely than the conventional spatial resolution of a power of 2, the code amount can be visually adjusted with little deterioration. According to the conventional decoding methods disclosed in Patent Documents 1 and 2, it is possible to obtain an image of arbitrary resolution by any decoding method, but the code to be transmitted or recorded is a power unit of 2. Thus, fine adjustment of the code amount is still impossible.

  Furthermore, since aliasing interference components (aliasing distortion) are mixed in the signal during the DWT downsampling process, it is necessary to prepare a plurality of removal filters having different characteristics. This aliasing interference component is a component that is removed by adding the low frequency component and the high frequency component, but if the low-resolution image is restored in a situation where the high frequency component is not transmitted, the aliasing interference component is not removed and the signal is removed. Appears as distortion.

  This will be described with reference to FIG. In FIG. 33, it is expressed as a one-dimensional signal for simplicity. FIG. 33A shows a low frequency component (L) created by downsampling. A region I indicated by diagonal lines below a half band (1/2 fs) of the Nyquist frequency in FIG. 33A indicates that aliasing interference generated by downsampling is mixed.

  The low frequency component shown in FIG. 33A is input to the input terminal 601 of the horizontal synthesizer 600 in FIG. 26, and the Nyquist frequency is upsampled from fs to fs ′ (= 2fs) by the upsampler 603. The Then, the folding interference changes as indicated by the hatched area II in FIG. Originally, the aliasing interference is removed by being combined with the high frequency component (H) input from the input terminal 602 by the adder 607 after passing through the combined H-LPF unit 605. The disturbance that appears when the high-frequency component is not transmitted is the component in the hatched area II shown in FIG.

  The present invention has been made in view of the above points, and an object of the present invention is to provide an image hierarchy encoding method and an image hierarchy decoding method having a code amount fine adjustment function based on spatial resolution.

  Another object of the present invention is to provide an image layer encoding method and an image layer decoding method capable of improving the obstruction caused by aliasing that occurs when high frequency components are not transmitted by a simple method so as not to stand out. It is in.

To achieve the former object, the image hierarchical encoding method of the present invention divides a two-dimensional image signal hierarchically into subbands, quantizes the subband signals obtained by the division, and then entropy codes. 1 is a picture hierarchical coding method for generating coded data and transmitting the coded data to a transmission line, a first step of octave-decomposing a two-dimensional image signal in a horizontal direction and a vertical direction; The second step to repeat the horizontal low and vertical high frequency signals obtained by the octave decomposition of the step in the octave decomposition only in the vertical direction, and the horizontal high and vertical low frequency signals obtained by the octave decomposition of the first step A third step for repeated octave decomposition only in the horizontal direction, a first signal subband obtained by the octave decomposition in the second step, and a third step. The horizontal high and vertical high frequency signals obtained by the octave decomposition in the first step are repeated in the horizontal and vertical directions so that they correspond to the subbands of the second signal obtained by the octave decomposition. Among the encoded data obtained by sequentially performing quantization and entropy encoding on the subband signals respectively obtained by the octave decomposition in the fourth step and the first to fourth steps, According to the transmission capacity of the transmission path, the encoded data corresponding to the subband signal with low resolution is transmitted to the transmission path in the order of the encoded data corresponding to the subband signal with high resolution, and the transmission capacity of the transmission path is insufficient. include a fifth step to stop the transmission of the coded data you characterized at this point. According to the present invention, it is possible to adjust the code amount with a resolution finer than a power of 2.

  In order to achieve the above-described object, the image hierarchical decoding method of the present invention performs entropy decoding on encoded data that is hierarchically encoded and transmitted by the image hierarchical encoding method of the present invention. A hierarchical decoding method for decoding an original image signal having the same number of pixels as before encoding from demodulated data of a subband signal obtained by performing quantization and dequantization sequentially, and encoding in which transmission is stopped A first step of substituting a zero value into the demodulated data area corresponding to the data, and from the demodulated data into which the zero value has been substituted in the first step, the signal repeatedly subjected to octave decomposition only in the horizontal direction The second step of synthesizing the vertical low frequency signal, and the horizontal low frequency / vertical high frequency signal from the signal repeatedly octave-decomposed only in the vertical direction among the demodulated data substituted with zero values in the first step And a fourth step of synthesizing a horizontal high-frequency / vertical high-frequency signal from a signal that is repeatedly octave-decomposed in the horizontal and vertical directions among the demodulated data into which zero values are substituted in the first step. And a fifth step of restoring the original image signal having the same number of pixels as that before encoding by performing horizontal and vertical synthesis processing on the signals obtained by the second to fourth steps. It is characterized by including.

  In order to achieve the latter object, the image hierarchical decoding method of the present invention includes a data area in which a zero value is substituted in the first step of demodulated data, and significant data obtained by octave decomposition of the demodulated data. When the data area of the significant data is on the low frequency side of the data area to which the zero value is substituted when combining the data area with When low-pass filtering characteristics with a cutoff frequency of 1/2 times the Nyquist frequency of the demodulated data are given and the data area of significant data is on the high frequency side of the data area to which the zero value is substituted, zero It is characterized in that a high-pass filtering characteristic having a cutoff frequency of 1/2 times the Nyquist frequency of the demodulated data is added to the synthesized signal obtained by synthesizing values and significant data.

  In this invention, in order to remove interference caused by folding when high-frequency components are not transmitted, a low-pass filtering characteristic or a high-pass filtering characteristic (that is, a frequency selection characteristic by a filter) is given in the synthesis process of each band. Therefore, it is sufficient to prepare only a 1/2 LPF or a 1/2 HPF as a filter, and it is possible to realize an interference removal configuration by folding that does not require a filter having a complicated characteristic.

  According to the present invention, since it is possible to finely adjust the code amount based on the spatial resolution, the code amount adjustment method based on the visual degradation degree is selected as the code amount adjustment based on the visual degradation degree as the code amount adjustment based on the quantization and the code amount adjustment based on the spatial resolution. can do. Therefore, there is an effect that it is possible to adjust the code amount that satisfies both the “request for the code amount of the transmission / recording medium” and the “request for the image quality of the viewer”, which had a strong trade-off relationship in the past.

  In addition, according to the present invention, aliasing interference (folding distortion) caused by a lack of a high frequency component can be avoided by using a simple 1/2 low-pass filter or 1/2 high-pass without requiring a complex filter. Can be removed with a filter.

  Next, the best mode for carrying out the invention will be described with reference to the drawings. The present invention is characterized in that the resolution can be easily changed by adding and deleting a coding unit, and a high-frequency component subjected to octave decomposition is further octave decomposed to realize a fine adjustment function of spatial resolution and code amount. is there.

  Conventionally, in a horizontal / vertical frequency space, when the spatial resolution is reduced by one step from the image of the band A in FIG. 15, the band E has an area ratio of 1/4. On the other hand, in one embodiment of the present invention, the division filter is configured so that five different bands from band A to band E can be changed in stages as shown in FIG.

  Specifically, as shown in FIG. 17, the HL1 component of the horizontal high band and the vertical low band is divided into four parts H1 to H4. In addition, the LH1 component of the horizontal low band and the vertical high band is divided into four to V1 to V4. Further, the S1 component is taken as the HH1 component of the horizontal high region and the vertical high region corresponding to the bands of H1 and V1, S21, S22, and S23 are taken as the components corresponding to H2 and V2, and the components corresponding to H3 and V3 are taken. S31, S32, and S33 are taken, and S41, S42, S43, S44, and S45 are taken as components corresponding to H4 and V4. However, in consideration of the occurrence of reversal between the high frequency and the low frequency at each octave decomposition, it is necessary to rearrange the bands as shown in FIG. 17 after dividing each band as shown in FIG.

  Further, when the high frequency component is not transmitted as a result of the code amount adjustment, the aliasing interference appears in the shaded area in FIG. 33B as described above. Accordingly, a low-pass filtering characteristic by a 1/2 low-pass filter (1 / 2LPF) shown in FIG. 19 as a folding interference elimination filter, that is, a low-pass filter (LPF) having a cutoff frequency of 1/2 times the Nyquist frequency. By adding to the folding interference, it is possible to eliminate the interference in the hatched area III shown in FIG. 19 and visually reduce the interference.

  However, when the band is finely adjusted as in the band dividing unit according to the embodiment of the present invention, it is conceivable to provide the filter characteristic of the aliasing elimination filter after restoring the image, but a complicated filter is prepared. There is a need. For example, as an example of the present invention, consider a case where the horizontal band is divided into bands A to E as shown in FIG. In FIG. 20A, the band E is a basic band. In this band E, the bands are expanded in the order of bands D, C, B, and A in accordance with the amount of code that can be transmitted.

  Here, the ◯ mark at the upper right of each band in FIG. 20A is a marker indicating the high frequency side of each band. At this time, if the filter characteristics of the aliasing removal filter corresponding to the transferred band are given after the image restoration, it is necessary to prepare four types of filters indicated by broken lines in FIG. .

  Accordingly, a configuration is considered in which the filter characteristic of the aliasing elimination filter is added in the synthesis process of the bands A, B, C, and D that have been transmitted after being divided. In the synthesis process of each band, first, the band D and the band C are synthesized as shown in FIG. 20 (d), and then the band A and the band B are synthesized as shown in FIG. 20 (c). Bands A, B, C, and D are combined as shown in FIG. 20B, and finally, all the bands are combined as shown in FIG.

  Accordingly, when the transmission up to the band D is performed, the interference is generated on the Takashiro side (the circle side) of the band D in the state shown in FIG. 20D, and therefore, as shown in FIG. A low-pass filter characteristic of a low-pass filter (hereinafter referred to as ½ LPF) having a frequency that is ½ the frequency as a cutoff frequency may be added to the synthesized signal.

  Further, when up to the band C is transmitted, a return interference occurs on the high band side (circle side) of the band C in the state of FIG. 20B, so as shown in FIG. A high-pass filter characteristic of a high-pass filter (hereinafter referred to as 1 / 2HPF) having a cutoff frequency of 1/2 times the Nyquist frequency may be added to the synthesized signal.

  Further, when the transmission up to the band B is performed, a return interference occurs on the high band side (circle side) of the band B in the state of FIG. 20 (c), so that 1 as shown in FIG. 20 (g). A high pass filtering characteristic of / 2HPF may be added to the synthesized signal. When only the band E is transmitted, a low-pass filtering characteristic of 1/2 LPF may be added to the synthesized signal in the state shown in FIG.

  As described above, it is only necessary to prepare 1 / 2LPF or 1 / 2HPF as a filter by adding a frequency selection characteristic by a filter in the synthesis process of each band, and there is a configuration that does not require a filter having a complicated characteristic. Realized. In particular, for 1/2 HPF, it is only necessary to subtract the output of 1/2 LPF from the original signal, and as a result, it is possible to realize an aliasing removal filter with only 1/2 LPF.

(First Example of Hierarchical Coding Method)
Next, a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a flowchart of a first embodiment of an image hierarchical encoding method according to the present invention, and FIG. 2 is a block diagram of an image hierarchical encoding apparatus corresponding to the first embodiment of the image hierarchical encoding method according to the present invention. . The first embodiment of the encoding method of the present invention is an embodiment in which the DWT unit 102 of FIG. 21 performs the subband division method of FIG. 18. First, the horizontal / vertical decomposer 700-1 of FIG. The image signal input to 701-1 is subjected to octave decomposition into components LL1, LH1, HL1, and HH1 (step S100 in FIG. 1).

  The horizontal / vertical decomposition in step S100 is realized by performing the processing described with reference to the horizontal / vertical decomposer 700 in FIG. 27 in the horizontal / vertical decomposer 700-1 in FIG. As a result, the image signal input from the input terminal 701-1 in FIG. 2 is octave-decomposed into LL1, LH1, HL1, and HH1, and output from the output terminals 702-1, 703-1, 704-1, and 705-1. Is done.

  Next, the LL1 component extracted in step S100 is subjected to horizontal / vertical decomposition and octave decomposition into LL2, LH2, HL2, and HH2 (step S101 in FIG. 1). The horizontal / vertical decomposition processing in step S101 is performed by the horizontal / vertical decomposition device 700-2 in FIG. 2, and the processing described for the horizontal / vertical decomposition device 700 in FIG. 27 is performed. As a result, in the horizontal / vertical decomposer 700-2, the LL1 component input from the input terminal 701-2 is octave-decomposed into LL2, LH2, HL2, and HH2, and output terminals 702-2, 703-2, and 704- 2 705-2.

  Further, the LH1 component extracted in step S100 is subjected to vertical four decomposition and octave decomposition into V1, V2, V3, and V4 (step S102 in FIG. 1). The vertical four decomposition process in step S102 can be performed using the vertical decomposer 300 shown in FIG. This operation is handled by the vertical four decomposer 1800 of FIG.

  FIG. 3 shows a block diagram of an example of the vertical four decomposer 1800 described above. The LH1 component coming from the input terminal 1801 in FIG. 3 is input to the input terminal 301-1 of the first vertical decomposer 300-1 and is octave decomposed in the vertical direction. Of the signals subjected to octave decomposition by the vertical decomposer 300-1, the vertical low frequency signal is output from the output terminal 306-1 and the vertical high frequency signal is output from the output terminal 307-1. The vertical low-frequency signal output from the output terminal 306-1 is input to the input terminal 301-2 of the second vertical decomposer 300-2, and is subjected to octave division in the vertical direction. Of the signals subjected to the octave decomposition by the vertical decomposer 300-2, the vertical low frequency signal is output from the output terminal 306-2, and the vertical high frequency signal is output from the output terminal 307-2.

  As a result, the vertical low frequency signal output from the output terminal 306-2 is output to the outside from the output terminal 1802 of the vertical 4 decomposer 1800 as a fourth vertical high frequency signal (V4). Further, the vertical high-frequency signal output from the output terminal 307-2 is output to the outside from the output terminal 1803 as a third vertical high-frequency signal (V3).

  Similarly, the vertical high-frequency signal output from the output terminal 307-1 is input to the input terminal 301-3 of the vertical decomposer 300-3, where it is octave decomposed in the vertical direction. As a result, the signal output from the output terminal 306-3 of the vertical decomposer 300-3 is output to the outside from the output terminal 1804 of the vertical four decomposer 1800 as the first vertical high frequency signal (V1). The signal output from the output terminal 307-3 of the vertical decomposer 300-3 is output from the output terminal 1805 to the outside as the second vertical high frequency signal (V2).

  Returning to FIG. 1 again, the HL1 component extracted in step S100 of FIG. 1 is horizontally decomposed into four parts and octave decomposed into H1, H2, H3, and H4 (step S103 of FIG. 1). The horizontal decomposition process in step S103 can be performed using the horizontal decomposition unit 400 shown in FIG. This operation is processed by the horizontal four decomposer 1700 of FIG.

  FIG. 4 shows a block diagram of an example of the horizontal four decomposer 1700 described above. The HL1 component coming from the input terminal 1701 in FIG. 4 is input to the input terminal 401-1 of the first horizontal decomposer 400-1 and is subjected to octave decomposition in the horizontal direction. Of the signals subjected to octave decomposition by the horizontal resolver 400-1, a horizontal low frequency signal is output from the output terminal 406-1, and a horizontal high frequency signal is output from the output terminal 407-1.

  The horizontal low-frequency signal output from the output terminal 406-1 is input to the input terminal 401-2 of the second horizontal decomposer 400-2, where it is octave decomposed in the horizontal direction. Of the signals subjected to octave decomposition by the horizontal resolver 400-2, a horizontal low frequency signal is output from the output terminal 406-2, and a horizontal high frequency signal is output from the output terminal 407-2. As a result, the signal output from the output terminal 406-2 is output to the outside from the output terminal 1702 of the horizontal 4 decomposer 1700 as the fourth horizontal high frequency signal (H4). The signal output from the output terminal 407-2 is output from the output terminal 1703 to the outside as a third horizontal high frequency signal (H3).

  Similarly, the horizontal high-frequency signal output from the output terminal 407-1 is input to the input terminal 401-3 of the horizontal decomposer 400-3, where it is octave decomposed in the horizontal direction. Of the signals subjected to octave decomposition by the horizontal decomposer 400-3, the horizontal high-frequency signal output from the output terminal 406-3 is output from the output terminal 1704 of the horizontal four decomposer 1700 as the first horizontal high-frequency signal (H1). The horizontal high-frequency signal output to the outside and output from the output terminal 407-3 is output to the outside from the output terminal 1705 as the second horizontal high-frequency signal (H2).

  Returning to FIG. 1 again, the HH1 component extracted in step S100 in FIG. 1 is the components S1, S21, S22, S23, S31, S32, S33, S41, S42, S43, S44, S45 described above. Is decomposed into octaves (step S104 in FIG. 1). The processing in step S104 is performed by the HH decomposer 1900 in FIG.

  FIG. 5 shows a block diagram of an example of the HH decomposer 1900 described above. In FIG. 5, the HH1 component coming from the input terminal 1901 is input to the input terminal 701-3 of the first horizontal / vertical decomposer 700-3 described with reference to FIG. Horizontal low-frequency / vertical low-frequency signal, horizontal low-frequency / vertical high-frequency signal, horizontal high-frequency / vertical low-frequency signal, horizontal high-frequency / vertical high-frequency signal obtained by octave decomposition by horizontal / vertical decomposer 700-3 The signals are output separately from output terminals 702-3, 703-3, 704-3, and 705-3.

  The horizontal low-frequency / vertical low-frequency signal output from the output terminal 702-3 is input to the input terminal 701-4 of the second horizontal / vertical decomposer 700-4, and again subjected to horizontal / vertical octave decomposition. . Signals obtained by octave decomposition by the horizontal / vertical decomposer 700-4 are output from output terminals 702-4, 703-4, 704-4, and 705-4. The signals output from the output terminals 702-4, 703-4, 704-4, and 705-4 are output from the output terminals 1902, 1903, 1904, and 1905 of the HH decomposer 1900 as signal components S43, S42, S44, and S32. Is output.

  Also, the horizontal low-frequency / vertical high-frequency signal output from the output terminal 703-3 of the horizontal / vertical decomposer 700-3 in FIG. 5 is input to the input terminal 401 of the horizontal decomposer 400 having the configuration shown in FIG. Octave decomposition is performed in the horizontal direction and output from the output terminals 406 and 407. The signals output from the output terminals 406 and 407 are output to the outside from the output terminals 1906 and 1907 of the HH decomposer 1900 as signal components S41 and S31.

  Further, the horizontal high frequency / vertical low frequency signal output from the output terminal 704-3 of the horizontal / vertical decomposer 700-3 in FIG. 5 is input to the input terminal 301 of the vertical decomposer 300 having the configuration shown in FIG. Octave decomposition is performed in the vertical direction and output from the output terminals 306 and 307. The signals output from the output terminals 306 and 307 are output to the outside from the output terminals 1908 and 1909 of the HH decomposer 1900 as signal components S45 and S33.

  Further, the horizontal high-frequency / vertical high-frequency signal output from the output terminal 705-3 of the horizontal / vertical decomposer 700-3 in FIG. 5 is supplied to the third horizontal / vertical decomposer 700-5 having the configuration shown in FIG. The signal is input to the input terminal 701-5 and is subjected to octave decomposition in the horizontal and vertical directions, and is output from the output terminals 702-5, 703-5, 704-5, and 705-5. The signals output from the output terminals 702-5, 703-5, 704-5, 705-5 are externally output from the output terminals 1910, 1911, 1912, 1913 of the HH decomposer 1900 as signal components S1, S23, S21, S22. Is output. With the above operation, the processing of FIG. 1 is completed, and the signal shown in FIG. 18 is obtained.

(Second Embodiment of Hierarchical Coding Method)
Next, a second embodiment of the encoding method of the present invention will be described with reference to the drawings. FIG. 6 shows a flowchart of a second embodiment of the image hierarchical encoding method according to the present invention. In the present embodiment, for example, in the code output processor 105 of FIG. 21, after all the encoding is completed, the code amount is adjusted by changing the resolution and transmitted to the desired transmission line. It is the image hierarchical encoding method regarding the method.

  In the conventional octave decomposition means, the code amount can be adjusted by changing from band A to band E as shown in FIG. In this case, it is realized by not transmitting LH1, HL1, and HH1 among the signals shown in FIG. However, as described above, the resolution can be adjusted only in units of powers of 2.

  On the other hand, according to the dividing method shown in FIG. 17 according to the present embodiment, first, signal components LL2, LH2, HL2, and HH2 are transmitted according to step S201 of FIG. Next, in the code output processor 105 in FIG. 21, it is determined whether or not the capacity for transmitting the data in the band D remains after all the encoding is completed (step S202 in FIG. 6). If it is determined in step S202 that data in band D cannot be transmitted, the process is terminated. In this case, up to band E in FIG. 16 can be restored on the receiving side.

  On the other hand, if it is determined in step S202 that band D data can be transmitted, signal components H1, S1, and V1 are transmitted (step S203 in FIG. 6). Subsequently, it is determined whether or not the capacity for transmitting data in band C remains (step S204 in FIG. 6). If it is determined in step S204 that data in band C cannot be transmitted, the process is terminated. In this case, up to band D in FIG. 16 can be restored on the receiving side.

  On the other hand, if it is determined in step S204 that band C data can be transmitted, signal components H2, S21, S22, S23, and V2 are transmitted (step S205 in FIG. 6). Subsequently, it is determined whether or not a capacity for transmitting data in band B remains (step S206 in FIG. 6). If it is determined in step S206 that band B data cannot be transmitted, the process ends. In this case, the reception side can restore up to band C in FIG.

  On the other hand, if it is determined in step S206 that band B data can be transmitted, signal components H3, S31, S32, S33, and V3 are transmitted (step S207 in FIG. 6). Subsequently, it is determined whether or not the capacity for transmitting data of band A remains (step S208 in FIG. 6). If it is determined in step S208 that data in band A cannot be transmitted, the process ends. In this case, the reception side can restore up to band B in FIG.

  On the other hand, if it is determined in step S208 that band A data can be transmitted, signal components H4, S41, S42, S43, S44, S45, and V4 are transmitted (step S209 in FIG. 6).

  As described above, according to the present embodiment, data that has been repeatedly octave-decomposed only in the horizontal direction, data that has been repeatedly octave-decomposed only in the vertical direction, and data that has been repeatedly octave-decomposed in the horizontal and vertical directions are converted into resolutions. Since the transmission is performed in ascending order and the processing is terminated when the transmission capacity is insufficient, that is, the transmission is stopped, the code amount can be adjusted with a finer resolution than the power of 2.

(First Embodiment of Hierarchical Decoding Method)
Next, an embodiment of the image layer decoding method according to the present invention will be described. FIG. 7 is a flowchart of a first embodiment of the image hierarchy decoding method according to the present invention, and FIG. 8 is a block diagram of an image hierarchy decoding apparatus corresponding to the first embodiment of the image hierarchy decoding method according to the present invention. . The first embodiment of the decoding method of the present invention is an embodiment for restoring the octave-decomposed signal in the IDWT unit 204 of FIG. 22 when the encoded data of the resolution band that fits in the transmission capacity is transmitted.

  First, a zero value is set in the data area that has not been transmitted to the receiving side in the processing of the flowchart shown in FIG. 6 (step S300 in FIG. 7). That is, the encoded data obtained by quantizing and entropy encoding the subband signal is sequentially transmitted to the transmission path from the encoded data corresponding to the subband signal having a low resolution according to the flowchart of FIG. Since transmission is stopped when the transmission capacity of the transmission path becomes insufficient, not all encoded data is transmitted depending on the transmission capacity. In step S300, a zero value is set in the data area of the demodulated data of the subband signal corresponding to the encoded data that has not been transmitted due to the lack of transmission capacity.

  Next, among the demodulated data of the subband signal, LL1 is restored from LL2, LH2, HL2, and HH2 by the first horizontal / vertical combiner 900-1 in FIG. 8 (step S301 in FIG. 7). The first horizontal / vertical synthesizer 900-1 performs the operation described above with reference to FIG. 29, and LL2, LH2 is connected to each of the input terminals 901-1, 902-1, 903-1, and 904-1 of FIG. , HL2 and HH2 are input, and the restored LL1 component is output from the output terminal 905-1.

  Further, the LH1 component is restored by performing the vertical four synthesis on the signal components V4, V3, V1, and V2 obtained by dividing the signal whose zero value is set in the untransmitted component in step S300 into four (step S302 in FIG. 7). ). The vertical 4 composition in step S302 is performed by the vertical 4 composition unit 2200 in FIG.

  FIG. 9 shows a block diagram of an example of the vertical 4 synthesizer 2200 shown in FIG. In FIG. 9, the V4, V3, V1, and V2 components are input to input terminals 2201, 2202, 2203, and 2204 of the vertical 4 synthesizer 2200, respectively. The V4 component input from the input terminal 2201 is input to the input terminal 501-1 of the first vertical synthesizer 500-1. Further, the V3 component input from the input terminal 2202 is input to the input terminal 502-1 of the first vertical synthesizer 500-1. The first vertical synthesizer 500-1 outputs the vertical high-frequency signal restored from the output terminal 508-1 by vertically synthesizing the input signals V4 and V3 by the operation described with reference to FIG.

  On the other hand, the V1 component input from the input terminal 2203 is input to the input terminal 501-2 of the second vertical synthesizer 500-2 in FIG. The V2 component input from the input terminal 2204 is input to the input terminal 502-2 of the second vertical synthesizer 500-2. The second vertical synthesizer 500-2 outputs the vertical high-frequency signal restored from the output terminal 508-2 by vertically synthesizing the input signals V1 and V2 by the operation described with reference to FIG.

  The vertical high-frequency signal output from the output terminal 508-1 of the first vertical synthesizer 500-1 is input to the input terminal 501-3 of the third vertical synthesizer 500-3, and the second vertical synthesizer 500. -2 output terminal 508-2 is input to the input terminal 502-3 of the third vertical synthesizer 500-3. The third vertical synthesizer 500-3 outputs the LH1 component restored from the output terminal 508-3 by vertically synthesizing the two-input vertical high-frequency signal by the operation described with reference to FIG. This LH1 component is output from the output terminal 2205 of the vertical 4 synthesizer 2200 to the outside.

  Referring back to FIG. 7 again, in step S300, the signal component H4, H3, H1, and H2 of the signal in which the zero value is set as the non-transmission component in step S300 is horizontally four-combined to generate the HL1 component. Restoration is performed (step S303 in FIG. 7). The horizontal 4 composition in step S303 is performed by the horizontal 4 composition unit 2100 in FIG.

  FIG. 10 shows a block diagram of an example of the horizontal 4-synthesizer 2100 shown in FIG. In FIG. 10, H4, H3, H1, and H2 components are input to input terminals 2101, 2102, 2103, and 2104 of the horizontal 4 synthesizer 2100, respectively. The H4 component coming from the input terminal 2101 is inputted to the input terminal 601-1 of the first horizontal synthesizer 600-1, and the H3 component coming from the input terminal 2102 is inputted to the first horizontal synthesizer 600-1. Input to the input terminal 602-1. The first horizontal synthesizer 600-1 is configured as described above with reference to FIG. 26. The first horizontal synthesizer 600-1 horizontally synthesizes the input H4 component and H3 component and outputs a restored horizontal high-frequency signal from the output terminal 608-1. .

  10 is input to the input terminal 601-2 of the second horizontal synthesizer 600-2, and the H2 component input from the input terminal 2104 is the second horizontal synthesizer. Input to the input terminal 602-2 of 600-2. The second horizontal synthesizer 600-2 is configured as described above with reference to FIG. 26, and horizontally synthesizes the input H1 component and H2 component and outputs a horizontal high-frequency signal restored from the output terminal 608-2. To do.

  The horizontal high-frequency signal output from the output terminal 608-1 of the first horizontal synthesizer 600-1 is input to the input terminal 601-3 of the third horizontal synthesizer 600-3, and the second horizontal synthesizer 600. -2 output terminal 608-2 is input to the input terminal 602-3 of the third horizontal synthesizer 600-3. The third horizontal synthesizer 600-3 has the configuration shown in FIG. 26, and horizontally synthesizes the two-input horizontal high-frequency signal and outputs the restored HL1 component from the output terminal 608-3. This HL1 component is output from the output terminal 2105 of the horizontal 4 synthesizer 2100 to the outside.

  Returning to FIG. 7 again, in parallel with the processing in steps S301 to S303, the HH1 component is restored from the signal in which the zero value is set in the untransmitted component in step S300 (step S304 in FIG. 7). The synthesis of the HH1 component in step S304 is performed by the HH synthesizer 2300 in FIG.

  FIG. 11 shows a block diagram of an example of the HH combiner 2300 shown in FIG. In FIG. 11, S43, S42, S44, and S32 components respectively input to input terminals 2301, 2302, 2303, and 2304 are input terminals 901-3, 902-3 of the first horizontal / vertical synthesizer 900-3. A signal that is input to 903-3 and 904-3 and restored by the configuration and operation described above with reference to FIG. 29 is output from the output terminal 905-3.

  Also, the components S41 and S31 input to the input terminals 2305 and 2306 of the HH combiner 2300 in FIG. 11 are input to the input terminals 601 and 602 of the horizontal combiner 600, respectively, and the configuration and operation shown in FIG. The signal restored from the output terminal 608 is output by the horizontal synthesis.

  Also, the components S45 and S33 input from the input terminals 2307 and 2308 of the HH combiner 2300 in FIG. 11 are input to the input terminals 501 and 502 of the vertical combiner 500, respectively, and the configuration and operation shown in FIG. The signal restored from the output terminal 508 is output by the vertical synthesis.

  Furthermore, the S1, S23, S21, and S22 components respectively input to the input terminals 2309, 2310, 2311, and 312, of the HH combiner 2300 in FIG. 11 are input to the input terminal 901 of the second horizontal / vertical combiner 900-4. 4, 902-4, 903-4, and 904-4, and a signal restored by the configuration and operation described above with reference to FIG. 29 is output from the output terminal 905-4.

  The output terminal 905-3 of the first horizontal / vertical synthesizer 900-3 in FIG. 11, the output terminal 608 of the horizontal synthesizer 600, the output terminal 508 of the vertical synthesizer 500, and the second horizontal / vertical synthesizer 900-4. The restored signal output from each of the output terminals 905-4 is input to the input terminals 901-5, 902-5, 903-5, and 904-5 of the third horizontal / vertical combiner 900-5. Thus, a signal restored by the configuration and operation described with reference to FIG. 29 is output from the output terminal 905-5. The signal output from the output terminal 905-5 of the third horizontal / vertical synthesizer 900-5 is output to the outside as an HH1 component from the output terminal 2313 of the HH synthesizer 2300.

  Returning to FIG. 7 again, when the processes in steps S301 to S304 are completed, horizontal and vertical composition is finally performed to restore the image (step S305 in FIG. 7). The processing in step S305 is performed by the second horizontal / vertical synthesizer 900-2 in FIG. The second horizontal / vertical synthesizer 900-2 in FIG. 8 includes the LL1 component output from the output terminal 905-1 of the first horizontal / vertical synthesizer 900-1 input to the input terminal 901-2, The LH1 component output from the output terminal 2205 of the vertical 4 synthesizer 2200 input to the input terminal 902-2 and the HL1 component output from the output terminal 2105 of the horizontal 4 synthesizer 2100 input to the input terminal 903-2. The HH1 component output from the output terminal 2313 of the HH combiner 2300 input to the input terminal 904-2 is horizontally and vertically combined by the configuration and operation described above with reference to FIG. 2 to output a restored image signal. With the operation described above, image restoration is completed when data in a resolution band that fits in the transmission capacity according to the present invention is transmitted.

(Second Embodiment of Hierarchical Decoding Method)
Next, a second embodiment of the decoding method of the present invention will be described with reference to the drawings. FIG. 12 is a flowchart of a second embodiment of the image hierarchy decoding method according to the present invention, and FIG. 13 is a block diagram of an image hierarchy decoding apparatus corresponding to the second embodiment of the image hierarchy decoding method according to the present invention. . In the present embodiment, for example, the horizontal 4 synthesizer 2900 shown in FIG. 13 is provided in the IDWT unit 204 shown in FIG. 22, and the aliasing interference (folding distortion) is removed from the input signal.

  In this embodiment, only the method of removing the folding interference in the horizontal bands (H1, H2, H3, H4) obtained by dividing the HL1 component into four parts by the horizontal four synthesizer 2900 of FIG. 13 will be described. However, the LH1 component and the HH1 Similarly, the component aliasing elimination method can be realized by inserting ½ LPF or ½ HPF every time each component is synthesized. For example, in 1/2 LPF, the values of seven tap coefficients are represented by [0.017, -0.085, 0.193, 0.75, 0.193, -0.085, 0.017]. A 7-tap digital filter can be used.

  13 is the same as the horizontal 4 synthesizer 2100 shown in FIG. 10 except that the zero value selectors 2905, 2906, 2907, and 2914, 1/2 (H-LPF) 2910, and 1/2 (H-HPF) 2908 and 2912 are used. And selectors 2909, 2911, and 2913 are added. In the following description, the description of the same function as that of the horizontal 4-synthesizer 2100 in FIG. 10 is omitted. Band A corresponds to the H4 component, band B corresponds to the H3 component, band C corresponds to the H2 component, and band D corresponds to the H1 component.

  First, the IDWT unit 204 checks whether the DWT coefficients of all the bands A, B, C, and D are transmitted, that is, whether there is a DWT coefficient up to the band A (step S401 in FIG. 12). If the DWT coefficients for the entire band have been transmitted, HL1 is synthesized by a normal synthesis operation (step S402 in FIG. 12). In the process of step S402, in the horizontal 4 synthesizer 2900 of FIG. 13, the zero value selectors 2905, 2906, 2907 and 2914 are operated so as to output the input signals as they are, and the selectors 2909, 2911 and 2913 are routed. This is done by operating to select the A side. In this case, the horizontal 4 synthesizer 2900 of FIG. 13 performs the same operation as the horizontal 4 synthesizer 2100 of FIG.

  Next, when it is determined in step S401 in FIG. 12 that the DWT coefficients of all bands up to bands A, B, C, and D have not been transmitted, the process proceeds to step S403, where bands B, C, and D are transmitted. That is, whether there is a DWT coefficient up to band B. If it is determined in step S403 in FIG. 12 that bands B, C, and D have been transmitted, the process proceeds to step S404, and the DWT coefficient for band A is set to zero.

  In the process of step S404 in FIG. 12, in the horizontal 4 synthesizer 2900 in FIG. 13, the zero value selector 2905 of the H4 component corresponding to the band A is used to set the zero value as the DWT coefficient to the first horizontal synthesizer 600-1. This is done by operating to input to the input terminal 601-1. Subsequently, band A and band B are synthesized, and a high-pass filtering characteristic of 1/2 HPF is given to the synthesized signal (step S405 in FIG. 12), and the high-frequency signal is restored by normal IDWT except for bands A and B. (Step S406 in FIG. 12).

  In the process of step S405, in the horizontal 4 synthesizer 2900 of FIG. 13, the selector 2909 is operated so as to select the path B side, and the high-pass filtering characteristic is applied by the 1/2 horizontal high-pass filter (1 / 2H-HPF device) 2908. Is output from the selector 2909 from the output signal of the output terminal 608-1 of the first horizontal synthesizer 600-1.

  Subsequently, the zero value selectors 2906, 2907, and 2914 in FIG. 13 output the input signals as they are, and the selectors 2911 and 2913 are operated so as to select the line A side, so that the process of step S406 is performed. As a result, up to the band B is restored, and the aliasing interference generated on the high band side of the band B is removed, and output from the horizontal 4 synthesizer 2900 output terminal 2915 in FIG.

  Next, when it is determined in step S403 of FIG. 12 that the DWT coefficients of the bands B, C, and D are not transmitted, the process proceeds to step S407, where the bands C and D are transmitted, that is, the band C Check if there is a DWT coefficient. When it is determined in Step S407 in FIG. 12 that the bands C and D are transmitted, after the DWT coefficients of the bands A and B are set to zero (Step S408 in FIG. 12), the band A is obtained by normal IDWT. , B, C, and D to restore the high frequency signal (step S409 in FIG. 12). Subsequently, the high-frequency signal restored in step S409 is provided with a high-pass filtering characteristic by 1 / 2HPF and output (step S410 in FIG. 12).

  In the process of step S408, in the horizontal 4 synthesizer 2900 in FIG. 13, the zero value selector 2905 supplied with the H4 component corresponding to the band A receives the zero value as the DWT coefficient from the first horizontal synthesizer 600-1. A zero value as a DWT coefficient is input to the input terminal 602-1 of the first horizontal synthesizer 600-1 from the zero value selector 2906 that is input to the input terminal 601-1 and supplied with the H3 component corresponding to the band B. This is done by the input process.

  Subsequently, corresponding to step S409 in FIG. 12, the zero value selector 2907 in FIG. 13 outputs the input signal H2 component as it is to the horizontal synthesizer 600-2, and the selectors 2909 and 2911 select the path A side. The signals output from the horizontal synthesizers 600-1 and 600-2 are supplied to the horizontal synthesizer 600-3 through the selectors 2909 and 2911 to be horizontally synthesized.

  Subsequently, corresponding to step S410 in FIG. 12, the signal output from the output terminal 608-3 of the third horizontal synthesizer 600-3 in FIG. 13 is converted into a 1/2 horizontal high-pass filter (1 / 2H-HPF). ) The selector 2913 is selected on the path B side so that the selector 2913 selects the signal to which the high-pass filtering characteristic is given by 2912, and the zero value selector 2914 is operated so as to output the input signal as it is. As a result, the output signal of the 1 / 2H-HPF 2912, that is, the signal from which the band C is restored and the aliasing interference generated on the high frequency side of the band C is removed passes through the selector 2913 and the zero value selector 2914 as it is. It is output from the output terminal 2915 of the horizontal 4 synthesizer 2900 of FIG.

  Next, when it is determined in step S407 in FIG. 12 that the DWT coefficients of the bands C and D are not transmitted, the process proceeds to step S411, and it is checked whether the DWT coefficients of the band D are transmitted. If the band D is transmitted in step S411, after setting the DWT coefficients of the bands A, B, and C to zero (step S412 in FIG. 12), the bands C and D are combined, and the 1/2 LPF is A low-pass filtering characteristic is added (step S413 in FIG. 12), and the high-frequency signal is restored by normal IDWT except for the bands C and D (step S414 in FIG. 12).

  In the horizontal 4 synthesizer 2900 of FIG. 13, corresponding to step S412 of FIG. 12, the first horizontal synthesizer receives the zero value as the DWT coefficient from the zero value selector 2905 to which the H4 component corresponding to the band A is supplied. A zero value as a DWT coefficient is input to the first horizontal synthesizer 600-1 from a zero value selector 2906 that is input to the input terminal 601-1 of 600-1 and supplied with an H3 component corresponding to the band B. A zero value as a DWT coefficient is input to the input terminal 602-2 of the second horizontal synthesizer 600-2 from the zero value selector 2907 that is input to the terminal 602-1 and supplied with the H2 component corresponding to the band C. Let them enter.

  Subsequently, corresponding to step S413 in FIG. 12, the horizontal combined signal of the bands C and D output from the output terminal 608-2 of the second horizontal combiner 600-2 in FIG. After giving a low-pass filtering characteristic by a low-pass filter (1 / 2H-LPF) 2910, the signal is input to the input terminal 602-3 of the horizontal synthesizer 600-3 through the selector 2911 operated to select the path B side. .

  Subsequently, corresponding to step S414 in FIG. 12, the selectors 2909 and 2913 in FIG. 13 are operated to select the path A side, and the zero value selector 2914 is operated to output the input signal as it is. As a result, the signal output from the horizontal synthesizer 600-3, that is, the signal from which the band D is restored and the aliasing interference generated on the high frequency side of the band D is removed is passed through the selector 2913 and the zero value selector 2914. It is output from the output terminal 2915 of the horizontal 4 synthesizer 2900 of FIG.

  If it is determined in step S411 in FIG. 12 that the DWT coefficient of band D is not transmitted, the process proceeds to step S415, and the high frequency signal is set to zero. Corresponding to the processing of step S415, in the horizontal 4 synthesizer 2900 of FIG. 13, the zero value selector 2914 outputs a zero value from the output terminal 2915 of the horizontal 4 synthesizer 2900.

  In this way, in the present embodiment, the processing of FIG. 12 with 1/2 HPF and 1/2 LPF added in step S303 of FIG. 7 is performed, and the HL1 component is divided into four parts using the horizontal 4-synthesizer 2900 of FIG. The folded interference in the horizontal band (H1, H2, H3, H4) can be eliminated.

  In addition, when removing the folding interference in the vertical band (V1, V2, V3, V4) obtained by dividing the LH1 component into four parts, it is the same as described with reference to FIGS. 12 and 13 using the vertical four combiner shown in FIG. Perform the process. The vertical 4 synthesizer 3000 shown in FIG. 14 outputs the V4, V3, and V2 components as they are or outputs the output signals of the zero value selectors 3005, 3006, and 3007 that output zero values, and the zero value selectors 3005 and 3006. A vertical synthesizer 500-1 for vertical synthesis, a vertical synthesizer 500-2 for vertical synthesis of the output signal of the zero value selector 3007 and the V1 component, 1 / 2V-HPF 2908 and 3012, and 1 / 2V-LPF 2910 , Selectors 3009, 3011, and 3013, a vertical synthesizer 500-3 that vertically synthesizes the output signals of the selectors 3009 and 3011, and a zero value selector 3014 that outputs the output signal of the selector 3013 as it is or outputs a zero value. It is composed of

  The selector 3009 selects one of the output signal of the vertical synthesizer 500-1 and the output signal of 1 / 2V-HPF 2908. The selector 3011 selects one of the output signal of the vertical synthesizer 500-2 and the output signal of the 1 / 2V-LPF 2910. The selector 3013 selects one of the output signal from the vertical synthesizer 500-3 and the output signal from the 1 / 2V-HPF 3012. Since the operation of the vertical 4 synthesizer 3000 is realized by replacing the horizontal processing in FIG. 13 with vertical processing, it can be easily inferred from FIG. 13 and will not be described in detail.

  The embodiment shown in FIGS. 12 to 14 is significant when the IDWT unit 204 combines the octave-decomposed zero-value substituted data area and the octave-decomposed data area with significant data. If the data area with correct data is on the lower frequency side than the data area where the zero value is assigned, immediately after combining the zero value and the significant data, the low-pass filtering characteristic by 1/2 LPF is given, and the significant data When a certain data area is on the higher frequency side than the data area to which the zero value is assigned, the high-pass filtering characteristic by 1/2 HPF is added immediately after combining the zero value and the significant data. (Folding distortion) can be removed.

  The present invention is not limited to the above embodiments and examples, and includes a computer program that causes a computer to execute the steps of the flowcharts of FIGS. 1, 6, 7, and 12. In this case, the computer program may be taken into the computer from the recording medium, or may be distributed and downloaded from the communication network to the computer.

It is a flowchart of 1st Example of the image hierarchy encoding method of this invention. It is a block diagram of the image hierarchy coding apparatus corresponding to 1st Example of the image hierarchy coding method of this invention. FIG. 3 is a block diagram of an example of a vertical four decomposer in FIG. 2. FIG. 3 is a block diagram of an example of a horizontal four decomposer in FIG. 2. It is a block diagram of an example of the HH decomposition device in FIG. It is a flowchart which shows the procedure of the code amount adjustment by the spatial resolution by 2nd Example of the image hierarchy encoding method of this invention. It is a flowchart of 1st Example of the image hierarchy encoding method of this invention. It is a block diagram of the image hierarchy decoding apparatus corresponding to 1st Example of the image hierarchy decoding method of this invention. It is a block diagram of an example of the vertical 4 synthesizer in FIG. It is a block diagram of an example of the horizontal 4 synthesizer in FIG. It is a block diagram of an example of the HH combiner in FIG. It is a flowchart which shows the process sequence at the time of inserting the aliasing interference reduction filter in the synthetic | combination process by 2nd Example of the image hierarchy decoding method of this invention. It is a block diagram of the image hierarchy decoding apparatus corresponding to 2nd Example of the image hierarchy decoding method of this invention. FIG. 13 is a diagram illustrating a vertical 4 synthesizer in which a folding interference reduction filter is inserted in the synthesis process according to the embodiment of FIG. 12. It is a figure explaining the spatial resolution which changes by the power-of-two unit. It is a figure explaining an example of the spatial resolution which changes in units other than the power of 2. It is a figure explaining when the band division by this invention is arranged in order of frequency. It is a figure explaining when the band division by this invention is arranged in the division order. It is a figure explaining the folding | turning interference reduction method by this invention. It is a figure explaining the effect | action of the folding | turning interference reduction filter in the synthetic | combination process by this invention. It is a block diagram which shows the basic composition of the encoding apparatus by DWT. It is a block diagram which shows the basic composition of the decoding apparatus by DWT. It is a block diagram of an example of the apparatus which carries out octave decomposition | disassembly perpendicularly | vertically. It is a block diagram of an example of the apparatus decomposed | disassembled octave horizontally. It is a block diagram of an example of the apparatus which synthesize | combines the signal by which the octave division | segmentation was carried out perpendicularly. It is a block diagram of an example of the apparatus which synthesize | combines the signal by which the octave decomposition | disassembly was carried out horizontally. It is a block diagram of an example of the apparatus which divides | segments otatab horizontally and vertically. It is a block diagram of an example of the apparatus disassembled twice. It is a block diagram of an example of the apparatus which decompress | restores the signal by which the octave decomposition | disassembly was carried out horizontally and vertically. It is a block diagram of an example of the apparatus which decompress | restores the signal decomposed | disassembled twice. It is a schematic diagram of the signal octave-decomposed horizontally and vertically. It is a schematic diagram of the signal decomposed | disassembled twice. It is a figure explaining the folding | turning disturbance at the time of combining only a low frequency component by making a high frequency component into zero.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 Image input processor 102 DWT unit 103 Quantizer 104 Entropy encoder 105 Code output processor 201 Code input processor 202 Entropy decoder 203 Inverse quantizer 204 IDWT unit 205 Image output processing unit 300 Vertical decomposer 302 Division V-LPF unit 303 Divided V-HPF unit 304,305,404,405 Downsampling unit 400 Horizontal decomposer 402 Divided H-LPF unit 403 Divided H-HPF unit 500 Vertical synthesizer 503,504,603,604 505 Combined V-LPF unit 506 Combined V-HPF unit 507,607 Adder 600, 600-1, 600-2, 600-3 Horizontal combiner 605 Combined H-LPF unit 606 Combined H-HPF unit 700, 700-1 , 700-2, 700-3, 7 0-4, 700-5 Horizontal / vertical decomposer 900, 900-1, 900-2, 900-3, 900-4, 900-5 Horizontal / vertical combiner 1700 Horizontal 4 decomposer 1800 Vertical 4 decomposer 1900 HH Decomposer 2100, 2900 Horizontal 4 synthesizer 2200, 3000 Vertical 4 synthesizer 2300 HH synthesizer 2905, 2906, 2907, 2914, 3005, 3006, 3007, 3014 Zero value selector 2908, 2912 1 / 2H-HPF device 2910 1 / 2H-LPF device 2909, 2911, 2913, 3009, 3011, 3013 Selector 3008, 3012 1 / 2V-HPF device 3010 1 / 2V-LPF device

Claims (3)

An image in which a two-dimensional image signal is hierarchically divided into subbands, subband signals obtained by the division are quantized, entropy-coded to generate encoded data, and the encoded data is transmitted to the transmission path A hierarchical encoding method comprising:
A first step of octave-decomposing the two-dimensional image signal in a horizontal direction and a vertical direction, and a horizontal low-frequency / vertical high-frequency signal obtained by the octave decomposition of the first step repeatedly octave-decomposing only in the vertical direction. Two steps,
A third step in which the horizontal high-frequency / vertical low-frequency signal obtained by the octave decomposition of the first step is repeatedly subjected to octave decomposition only in the horizontal direction;
The first sub-bands obtained by the octave decomposition of the second step correspond to the sub-bands of the second signal obtained by the octave decomposition of the third step. A fourth step in which the horizontal high-frequency / vertical high-frequency signal obtained by the octave decomposition of the step is repeatedly octave decomposed in the horizontal and vertical directions ;
Of the encoded data obtained by sequentially performing quantization and entropy encoding on the subband signals respectively obtained by the octave decomposition in the first to fourth steps, the transmission capacity of the transmission path According to the transmission data in the order of the encoded data corresponding to the subband signal having a low resolution to the encoded data corresponding to the subband signal having a high resolution, and the code when the transmission capacity of the transmission path is insufficient. A fifth step of stopping transmission of the digitized data;
Image hierarchical coding method, which comprises a. The demodulated data of the subband signal obtained by sequentially performing entropy decoding and inverse quantization on the encoded data transmitted by hierarchical encoding by the image hierarchical encoding method according to claim 1. An image hierarchical decoding method for decoding an original image signal having the same number of pixels as that before encoding,
A first step of substituting a zero value in the demodulated data area corresponding to the encoded data for which transmission has been suspended;
A second step of synthesizing a horizontal high-frequency / vertical low-frequency signal from a signal that is repeatedly octave-decomposed only in the horizontal direction among the demodulated data into which zero values are substituted in the first step;
A third step of synthesizing a horizontal low-frequency / vertical high-frequency signal from a signal that is repeatedly octave-decomposed only in the vertical direction among the demodulated data into which zero values are substituted in the first step;
A fourth step of synthesizing a horizontal high frequency / vertical high frequency signal from the signal repeatedly octave-decomposed in the horizontal direction and the vertical direction among the demodulated data into which the zero value is substituted in the first step;
A fifth step of restoring the original image signal having the same number of pixels as before encoding by performing horizontal and vertical synthesis processing on the signals obtained in the second to fourth steps. The image hierarchy decoding method characterized by the above-mentioned. When synthesizing the data area into which the zero value is substituted in the first step of the demodulated data and the data area with significant data obtained by octave decomposition of the demodulated data, the data area of the significant data is If it is on the low frequency side of the data area where the zero value is assigned, a frequency that is ½ times the Nyquist frequency of the demodulated data is cut from the synthesized signal obtained by synthesizing the zero value and the significant data. When a low-pass filtering characteristic is applied as an off-frequency, and the data area of the significant data is on the high frequency side of the data area to which the zero value is substituted, a synthesized signal obtained by synthesizing the zero value and the significant data respect, the image layer decoding side according to claim 2, wherein the benzalkonium be given a high-pass filter characteristics that the cutoff frequency 1/2 times the frequency of the Nyquist frequency of the demodulated data .

Images