JP3704644B2 - Image coding apparatus and method thereof, and image decoding apparatus and method thereof - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image encoding apparatus and method, an image decoding apparatus and method, and is suitable for application to, for example, an image providing system that provides an image.
[0002]
[Prior art]
In recent years, an image transmission device and an image reception device have been connected via a network, and the image
An image providing system that provides an image from an image transmitting apparatus to an image receiving apparatus is considered.
[0003]
[Problems to be solved by the invention]
By the way, in such an image providing system, for the purpose of protecting the copyright for the image to be provided, anyone can freely acquire it by embedding a digital watermark in the image to prevent unauthorized copying or encrypting the image. It has been proposed to make it possible, for example, to show only the person who purchases the image a key for decryption.
[0004]
However, in the method of embedding a digital watermark in the image in order to protect the copyright of the image, a part of the image data of the image is replaced with the data of the digital watermark and the image quality is deteriorated. It is unsuitable for etc.
[0005]
Also, in the method of encrypting the image, if the encryption is performed so that the outline of the image and the like of the image cannot be visually recognized at all, the image is not visible at all. It becomes difficult to provide easily.
[0006]
For this reason, recently, in an image providing system, encryption is performed so that an outline of a pattern or the like can be visually recognized to some extent on an image (hereinafter referred to as “scramble” in particular), and the image is actually purchased. It has been proposed that only a person can pass a key to unscramble and show an image.
[0007]
Further, in such an image providing system, when an image is provided, the image transmitting apparatus compresses and encodes the image data of the image by a JPEG (Joint Photographic Experts Group) method and transmits the image data to the image receiving apparatus. It has also been proposed to improve efficiency.
[0008]
Therefore, in such an image providing system, the image transmission device performs discrete cosine transform (DCT) on the image data, quantizes the obtained discrete cosine transform coefficient to generate a quantized coefficient, After the scramble process is performed on the quantized coefficient, the image data is compressed and encoded by performing a Huffman encoding process.
[0009]
Here, Huffman coding encodes a scrambled quantized coefficient, assigns a short code to the obtained code if the frequency of occurrence is high, and assigns a relatively long code to the code if the frequency of occurrence is low. Conversion process.
[0010]
However, in the image transmission apparatus, when the quantized coefficients are scrambled, the arrangement of the quantized coefficients changes randomly, so that the codes obtained by encoding the scrambled quantized coefficients are short. Compressed codes obtained by compressing and encoding image data because codes with relatively high occurrence frequency to which codes are assigned are greatly reduced, and codes with relatively low occurrence frequencies to which long codes are assigned are increased. The amount of data in the computerized data increases.
[0011]
For this reason, although the image providing system compresses and encodes image data, there is a problem in that the image data is transmitted at a compression rate much higher than a desired compression rate.
[0012]
The present invention has been made in consideration of the above points, and an object of the present invention is to propose an image encoding apparatus and method, an image decoding apparatus and method that can prevent a reduction in the compression rate of image data.
[0013]
[Means for Solving the Problems]
In order to solve such a problem, in the present invention, wavelet transform for recursively filtering low-frequency components is performed on image data to generate wavelet coefficients in subband units as filter coefficients, and the generated wavelet coefficients are Quantize to generate a quantized coefficient, encode the quantized coefficient with a predetermined encoding method to generate arithmetic code data, and then scramble to the specified part of the arbitrary subband unit of the arithmetic code data Performs processing, generates predetermined coding information related to the coding processing based on the arithmetic code data, and adds the scrambled arithmetic code data to the header data generated by storing the generated coding information To generate packet data.
[0014]
Therefore, after the quantization coefficient is encoded, the specified portion of the arbitrary subband unit of the arithmetic code data obtained by the encoding process is subjected to the scramble process, so that the image quality of the image based on the image data is improved. Deterioration can be reduced, and an increase in the amount of data when the quantized coefficients are encoded can be prevented.
[0015]
Further, in the present invention, the wavelet transform for recursively filtering the low-frequency component is applied to the image data, and the sub-band wavelet coefficients generated as filter coefficients are quantized, and then a predetermined encoding method is used. Generated by adding arithmetic code data that is scrambled to the specified part of any subband unit to the header data that stores the encoding information when arithmetic code data is generated by encoding Header data and scrambled arithmetic code data are separated from the received packet data, the scrambled arithmetic code data is descrambled to return to the original arithmetic code data, and the header data Based on the encoded information stored in the After generating a quantized coefficient by decoding with the method, the generated quantized coefficient is dequantized to generate a filter coefficient, and then the filter coefficient is filtered using a low-pass filter and a high-pass filter. Image data was generated.
[0016]
Therefore, with such a configuration, the image data encoding side has a reversible configuration with the configuration, and after encoding the quantized coefficient, the arithmetic code obtained by the encoding process is obtained. Since scramble processing is performed on a specified part of an arbitrary subband unit of data, image quality degradation can be reduced with respect to an image based on image data, and the amount of data increases when a quantization coefficient is encoded. This can be prevented.
[0017]
In FIG. 1, reference numeral 1 denotes an image providing system to which the present invention is applied as a whole, and an image transmitting apparatus 2 and an image receiving apparatus to which a compression encoding method called JPEG-2000, which is the next-generation still image international standard, is applied. 3 are connected via a network 4.
[0018]
In this case, in the image providing system 1, for example, when the image receiving device 3 requests the image transmitting device 2 to acquire an image via the network 4, the image transmitting device 2 sends the image data corresponding to the request to the JPEG. After compressing and encoding with the −2000 system, the image data is transmitted to the image receiving apparatus 3 via the network 4. When the image receiving apparatus 3 receives the compressed and encoded image data, the image data is decompressed and decoded with the JPEG-2000 system. It is made to become.
[0019]
In practice, as shown in FIG. 2, in the image transmission apparatus 2, when transmitting an image, for example, image data D1 of an image to be provided reproduced from an internal predetermined recording medium (not shown) is converted into a wavelet transform unit. 10
[0020]
The wavelet transform unit 10 performs wavelet transform on the image data D1 by filtering using a low-pass filter and a high-pass filter, and sends the obtained wavelet transform coefficient D2 to the quantization unit 11.
[0021]
The quantization unit 11 performs scalar quantization that divides the wavelet transform coefficient D2 given from the wavelet transform unit 10 by a predetermined quantization step size, and sends the obtained quantization coefficient D3 to the encoding unit 12. .
[0022]
The encoding unit 12 includes EBCOT (Embedded Coding with Optimized Truncation) defined by the JPEG-2000 system in units of a predetermined code block (code-block) with respect to the quantization coefficient D3 given from the quantization unit 11. Entropy coding called is performed, and arithmetic coding called MQ coding is performed on the obtained code, so that the obtained arithmetic code data D4 is sequentially sent to the rate control unit 13 in units of code blocks.
[0023]
Incidentally, MQ coding is learning type binary arithmetic coding defined by JBIG2 (Joint Bi-Level Image Coding Experts Group 2).
[0024]
The rate control unit 13 counts the code amount (that is, the data amount) of the arithmetic code data D4 sequentially given from the encoding unit 12 in units of code blocks, and a part or all of the arithmetic code data D4 for each code block. The code amount is controlled to be constant so that the target bit rate or compression coding rate is obtained, and the arithmetic code data D5 having a code amount corresponding to the target bit rate or compression coding rate obtained as a result is obtained. Is sent to the arithmetic code extraction unit 14 and the header generation unit 15.
[0025]
Here, in the image transmission device 2, the arithmetic code data D5 to be scrambled can be arbitrarily designated, and the arithmetic code extraction unit 14 performs arithmetic in code block units given from the rate control unit 13. Whether the code data D5 is arithmetic code data D5 in units of code blocks designated to be scrambled, or is arithmetic code data D5 in units of code blocks designated not to be scrambled Is added to the scramble processing unit 16.
[0026]
The scramble processing unit 16 is provided with public key data from a control unit (not shown), and an arithmetic code designated to perform scramble processing in the arithmetic code data D5 in units of code blocks supplied from the arithmetic code extraction unit 14 The data D5 is scrambled using the public key data, and the obtained scrambled data D6 is sent to the packet generator 17 and the arithmetic code in units of code blocks designated not to be scrambled The data D5 is sent to the packet generator 17 as it is.
[0027]
At this time, the header generation unit 15 is based on the arithmetic code data D5 in units of code blocks, the data length thereof, predetermined encoding information related to the encoding process, information indicating the scrambled arithmetic code data D5, and the like Is stored in a packet header defined by the JPEG-2000 system to generate header data D7, and the header data D7 is sent to the packet generator 17.
[0028]
Therefore, the packet generation unit 17 adds a plurality of scrambled data D6 in units of code blocks or a plurality of arithmetic code data D5 in units of code blocks provided from the scramble processing unit 16 to the header data D7 provided from the header generation unit 15. In addition, packet data is sequentially generated, the generated packet data is unified, and the obtained packet stream D8 is transmitted from the predetermined interface circuit 18 to the image transmission device 3 via the network 4.
[0029]
On the other hand, as shown in FIG. 3, in the image reception device 3, the packet stream D <b> 8 transmitted from the image transmission device 2 via the network 4 is taken into the packet decoding unit 21 via the predetermined interface circuit 20.
[0030]
The packet decoding unit 21 sequentially separates the fetched packet stream D8 into packet data, sends header data D7 included in the packet data to the header decoding unit 22, and scramble data D6 or arithmetic code included in the packet data The data D5 is sent to the descrambling processor 23.
[0031]
At this time, the descrambling processing unit 22 is provided with public key data from a control unit (not shown) and also shows arithmetic code data D5 subjected to scrambling processing obtained from the header decrypting unit 22 based on the header data D7. Information is given, and based on the information, the scrambled data D6 and the scrambled data D6 of the arithmetic code data D5 given from the packet decrypting unit 21 are selected and descrambled with the public key data. The original arithmetic code data D5 is sent to the arithmetic code decoding unit 24, and the arithmetic code data D5 that has not been scrambled at the time of transmission is sent to the arithmetic code decoding unit 24 as it is.
[0032]
Since the arithmetic code data D5 is sequentially supplied from the descrambling processing unit 22, the arithmetic code decoding unit 24 rearranges the sequential arithmetic code data D5 into a code block and sends it to the decoding unit 25.
[0033]
At this time, the header decoding unit 22 extracts the data length and encoding information stored in the packet header based on the packet data D7 given from the packet decoding unit 21, and according to the extracted data length and encoding information The decryption control data D10 is generated and sent to the decryption unit 25.
[0034]
Therefore, the decoding unit 25 performs arithmetic decoding processing on the arithmetic code data D5 in units of code blocks given from the arithmetic code decoding unit 24 based on the decoding control data D10 given from the header decoding unit 22. The entropy decoding process is performed, and the obtained quantized coefficient D11 is sent to the inverse quantization unit 26.
[0035]
The inverse quantization unit 26 inversely quantizes the quantization coefficient D11 given from the decoding unit 25, and sends the obtained wavelet transform coefficient D12 to the wavelet inverse transform unit 27.
[0036]
The wavelet inverse transform unit 27 performs wavelet inverse transform on the wavelet transform coefficient D12 given from the inverse quantizer 26 by filtering using a low-pass filter and a high-pass filter, and the obtained image data D13 is, for example, a predetermined display unit ( (Not shown), an image based on the image data D13 can be displayed on the display unit.
[0037]
Here, in the image transmission device 2, the wavelet transform unit 10 has a filter coefficient having a plurality of tap lengths, and performs filtering in the vertical direction of the image (hereinafter referred to as vertical filtering) and filtering in the horizontal direction ( Hereinafter, a low-pass filter and a high-pass filter used for horizontal filtering are provided, and a predetermined line buffer is provided.
[0038]
Then, as shown in FIG. 4, the wavelet transform unit 10 reads the image data D1 from the recording medium in units of line LI of the image SG based on the image data D1, stores it in the line buffer 30, and performs vertical filtering in the wavelet transform. When the data for the plurality of lines LI is stored as much as necessary, the data for the plurality of lines LI is subjected to vertical filtering, and the obtained data for the plurality of lines LI is subsequently subjected to horizontal filtering.
[0039]
As a result, as shown in FIGS. 5A and 5B, the wavelet transform unit 10 includes the first level lowest band subband LL1, the low and high band subband LH1, and the high and low band subbands having four bands. The wavelet transform coefficient D2 between the band HL1 and the highest band subband HH1 can be obtained.
[0040]
Here, the wavelet transform unit 10 obtains the wavelet transform coefficient D2 of the first level lowest band subband LL1, low high band subband LH1, high low band subband HL1, and highest band subband HH1. If specified, the wavelet transform coefficient D2 of these lowest band subband LL1, low and high band subband LH1, high and low band subband HL1, and highest band subband HH1 is continuously obtained, and the obtained lowest band The wavelet transform coefficient D2 of the subband LL1, the low and high frequency subband LH1, the high and low frequency subband HL1, and the highest frequency subband HH1 is sequentially sent to the quantization unit 11, and the wavelet transform coefficient is sent to the quantization unit 11 The quantization process for D2 is executed in parallel, and is designated from the image SG for one frame thereafter. Until obtaining all wavelet transform coefficients D2 of the first level to continue the series of processes.
[0041]
On the other hand, the wavelet transform unit 10 includes the lowest level subband LL2, the low and high frequency subband LH2, the high and low frequency subband HL2, and the highest frequency subband HH2 up to the second level as shown in FIG. 7A and 7B, as shown in FIGS. 7A and 7B, the lowest level subband LL1 of the first level, the low and high frequency subband LH1, and the high and low frequencies are specified. While continuously obtaining the wavelet transform coefficient D2 of the region subband HL1 and the highest region subband HH1, the obtained high region side low high region subband LH1, high region subband HL1, and highest region subband HH1 Are sequentially sent to the quantizing unit 11.
[0042]
In addition, the wavelet transform unit 10 sequentially stores the wavelet transform coefficients of the lowest level subband LL1 of the first level in the line buffer 30 in units of lines, and the line buffer 30 only requires vertical filtering in the wavelet transform. When the wavelet transform coefficients for a plurality of lines are stored, vertical filtering is performed on the wavelet transform coefficients for the plurality of lines, and then horizontal filtering is performed to obtain the lowest-level subband LL1 of the first level. A wavelet transform coefficient D2 of the lowest level subband LL2, the low and high frequency subband LH2, the high and low frequency subband HL2, and the highest frequency subband HH2 divided into the bands is obtained.
[0043]
Thereby, the wavelet transform unit 10 sends the wavelet transform coefficients D2 of the first level low and high frequency sub-band LH1, the high and low frequency sub-band HL1, and the highest frequency sub-band HH1 to the quantization unit 11 sequentially. On the other hand, the wavelet transform coefficient D2 of the second level lowest band subband LL2, low high band subband LH2, high low band subband HL2, and highest band subband HH2 is obtained and sent to the quantization unit 11. Thereafter, this series of processing is continued until all the wavelet transform coefficients D2 from the image SG for one frame to the designated second level are obtained.
[0044]
In this way, the wavelet transform unit 10 uses a small-capacity line buffer without using a large-capacity memory corresponding to the image SG for one frame, and uses a predetermined amount of data (or wavelet transform coefficients) in the line buffer. Each time D2) is accumulated, the wavelet transform can be executed sequentially and efficiently.
[0045]
Incidentally, as shown in FIGS. 8A and 8B, the wavelet transform unit 10 includes, for example, the lowest frequency subbands LL3 and LL4 up to the third level and the fourth level, and the low and high frequency subbands LH3 and LH4. And the lowest subband at a level that is one level lower than the desired level when the wavelet transform coefficient D2 of the high and low band subbands HL3 and HL4 and the highest band subbands HH3 and HH4 is specified. On the other hand, if the filtering process according to the filtering process as described above with reference to FIGS. 6A and 6B is executed, the wavelet transform coefficient D2 of the subband up to the desired level can be obtained.
[0046]
As shown in FIG. 9, the quantization unit 11 quantizes the wavelet transform coefficients D2 sequentially given from the wavelet transform unit 10 for each line of each subband, and encodes the obtained quantized coefficients D3. To send.
[0047]
As shown in FIGS. 10A and 10B, the encoding unit 12 has, for example, 64 quantization coefficients D3 for each subband given from the quantization unit 11 along the vertical direction indicated by the arrow a. When the number of samples h1 is reached, a code block consisting of the same number of samples h1 and h2 (that is, h1 and h2 are both 64 samples) in the vertical direction and in the horizontal direction orthogonal to the quantization coefficient D3 for each subband. Cut out and encode each cut out code block to generate arithmetic code data D4.
[0048]
10A and 10B show subbands up to the second level. As described above with reference to FIGS. 7A and 7B, the first level subbands and the second level subbands are shown in FIG. Since there is a time difference until the wavelet transform coefficient D2 with the level subband is obtained, the encoding unit 12 first determines the quantization coefficient of the first level subband that reaches the predetermined number of samples h1 along the vertical direction. When the quantization coefficient D3 of the second level subband reaches a predetermined number of samples h1 (64 samples) along the vertical direction, the first level subband is encoded. In addition to the quantization coefficient D3, the second-level subband quantization coefficient D3 is also encoded in units of code blocks.
[0049]
In this manner, the encoding unit 12 encodes all the subband quantization coefficients D3 obtained from the image for one frame in units of code blocks.
[0050]
Actually, as shown in FIG. 11 (A), the quantization coefficient D3 in units of code blocks has a maximum absolute value of 13 (for example, “1101” in binary representation) for each sample in the code block. Each has a coefficient value.
[0051]
Therefore, as illustrated in FIGS. 11B and 11C, the encoding unit 12 performs the encoding based on the quantization coefficient D3 in units of code blocks given from the quantization unit 11 when performing encoding. The absolute value of the quantization coefficient D3 is sequentially sliced at each bit from the least significant bit (LSB) to the most significant bit (MSB: Most Significant Bit), and “1” or “ An absolute value bit plane having a coefficient value of “0” (hereinafter referred to as an absolute value bit plane) is generated (in this case, four types of absolute value bit planes are generated), and the sign of each coefficient value ( ±) bit planes (hereinafter referred to as sign bit planes) are generated.
[0052]
Incidentally, in FIGS. 11A to 11C, for the sake of convenience, the size of the absolute value bit plane and the sign bit plane are represented as vertical and horizontal 4 bits, respectively. The size of each is 64 bits corresponding to 64 samples in the vertical and horizontal directions of the code block.
[0053]
Then, as shown in FIG. 12, the encoding unit 12 has, for example, a 4-bit height (vertical) and a 64-bit width (horizontal) for each block with respect to the absolute value bit plane. The stripe ST is divided into a plurality of stripes ST, and the coefficient values are sequentially scanned from the top to the bottom along the direction indicated by the solid arrow b and the dotted arrow c (solid arrow b). All the coefficient values in each stripe ST are scanned so as to shift from right to right (dotted arrow c), and the coefficient values are encoded.
[0054]
By the way, in entropy coding called EBCOT used in the coding unit 12, as a coding method for coefficient values in the absolute value bit plane, a significance pass, a refinement pass, and a cleanup pass are used. Three types of encoding pass called (Cleanup Pass) are defined.
[0055]
Also, in such entropy coding, when a significance pass, a refinement pass, and a cleanup pass coding pass are used, each of the coefficient values in each block of the absolute value bit plane is valid (significant) or A data table (hereinafter referred to as a state variable table) representing an invalid (non-significant) encoding state variable is used.
[0056]
On the other hand, the encoding unit 12 performs ZC (ZeroCoding), RLC (Run-Length Coding) while measuring statistics of arithmetic coding for each code block unit by MQ coding as arithmetic coding executed in the coding pass. ), SC (Sign Coding) and MR (Magnitude Refinement) are appropriately selected and executed. As a result, in the JPEG-2000 system in which arithmetic coding is performed within the coding pass of such entropy coding, there are a total of 19 types of contexts in all coding passes.
[0057]
Accordingly, as shown in FIG. 13, when the quantization coefficient D3 in units of code blocks has an absolute value of n-bit coefficient values from the least significant bit to the most significant bit, the encoding unit 12 has the most significant bit (n -1 bit) absolute value bit plane is encoded only by a cleanup pass while scanning coefficient values as described above with reference to FIG. 12, and each absolute value from the next n-2 to the least significant bit (0 bit) As described above with reference to FIG. 12, for each bit plane, three types of encoding processes are sequentially performed independently in the order of the significance pass, the refinement pass, and the cleanup pass for each absolute value bit plane while scanning the coefficient value. Execute.
[0058]
At this time, the encoding unit 12 stores the state variable table in the internal memory in advance, and when the encoding process is started with respect to the quantization coefficient D3 of one code block unit, the state variable table is stored from the memory. And all state variables are set as initial values, for example, “0” representing invalidity.
[0059]
Then, when one of the coefficient values is encoded, the encoding unit 12 changes the corresponding state variable on the state variable table to, for example, “1” indicating validity, and in this way, quantization of one code block unit Until the encoding process for the coefficient D3 is completed, the state variables that have become valid are not changed as they are, but only the state variables corresponding to the newly encoded coefficient values are sequentially changed and used.
[0060]
Incidentally, since the state variables in the state variable table are sequentially changed from invalid to valid each time the coefficient values are encoded in this way, it is determined whether or not the information on the effective digits (that is, the coefficient values) has already been encoded. It can be said that it represents a flag.
[0061]
Actually, when the absolute value bit plane is encoded by a significance path, the encoding unit 12 scans a coefficient value whose state variable is invalid, and among the coefficient values in the vicinity of 8 surrounding the coefficient value, If the state variable of at least one coefficient value is valid, the scanned coefficient value (that is, the state variable is invalid) is arithmetically encoded, and the arithmetically encoded coefficient value is “1”. Sometimes it is arithmetically encoded whether the corresponding code in the coded bitplane is “+” or “−”, thus signifying a single significant bitplane Encode with
[0062]
In addition, when the encoding unit 12 encodes the absolute value bit plane with the refinement pass, the encoding unit 12 has encoded with the previous significance path among the coefficient values for which the state variable in the absolute value bit plane is valid. A coefficient value that does not exist is arithmetically encoded, and in this way, the entire absolute value bit plane is encoded in the refinement pass.
[0063]
Furthermore, when the absolute value bit plane is encoded by the cleanup path, the encoding unit 12 does not encode the coefficient value whose state variable in the absolute value bit plane is invalid by the previous significance path. The coefficient value is arithmetically encoded, and when the arithmetically encoded coefficient value is “1”, it is continued whether the corresponding code of the encoded bit plane is “+” or “−”. Arithmetic encoding is performed, and in this way, the entire absolute value bit plane is encoded by a cleanup pass.
[0064]
However, when the encoding unit 12 encodes the absolute value bit plane of the most significant bit (n−1 bits) by the cleanup pass, all the state variables in the state variable table are initialized and become invalid. Therefore, the coefficient value that is “1” among the coefficient values in the absolute value bit plane is arithmetically encoded, and the consecutive number of “0” is detected for the coefficient value of “0”. For example, RLC is executed by MQ coding.
[0065]
Here, for the arithmetic code extraction unit 14, the arithmetic code data D5 to be scrambled can be specified alone. For example, as shown in FIGS. Sub-bands such as level low high band sub-band LH1, high low band sub-band HL1 and highest band sub-band HH1, second level low high band sub-band LH2, high low band sub-band HL2 and highest band sub-band HH2 It can also be specified in units.
[0066]
Therefore, when the subband to be scrambled is designated, the arithmetic code extraction unit 14 designates the designated first level low and high frequency subbands LH1, HL1 and HL1, and the highest frequency subband HH1, Information specifying that scramble processing is to be performed on all code block unit arithmetic code data D5 belonging to the second level low and high frequency subband LH2, high and low frequency subband HL2 and highest frequency subband HH2 is added. .
[0067]
For the arithmetic code extraction unit 14, all subbands and levels can be designated as subbands to be scrambled for one image, or a plurality of levels of subbands can be designated simultaneously. It is made like that.
[0068]
Incidentally, in the subbands up to a predetermined level, most of the energy constituting the image is concentrated in the low-frequency component, and as described above with reference to FIGS. Designates to scramble only subband LH1, high and low band subband HL1 and highest band subband HH1, and second level low and high band subband LH2, high and low band subband HL2, and highest band subband HH2. Instead, if it is specified that only the lowest subbands LL1 and LL2 are to be scrambled, the scramble strength can be increased accordingly.
[0069]
Then, the scramble processing unit 16 fixes the arithmetic code data D5 to the arithmetic code data D5 designated to be scrambled based on a common key block encryption method called a Blowfish method. The data is divided into long data strings and scrambled using the common key data for each data string.
[0070]
At this time, since the code amount is controlled by the rate control unit 13 so that the arithmetic code data D5 in units of code blocks has a constant data length, the scramble processing unit 16 uses the arithmetic code data D5 for scramble processing. The predetermined data string can be easily divided.
[0071]
In addition, the scramble processing unit 16 divides the arithmetic code data D5 into a fixed-length data string by a block encryption method and performs scramble processing using the common key data for each data string, thereby performing data after the scramble process. An increase in the amount can be almost certainly prevented.
[0072]
If the arithmetic code data D5 to be scrambled is arbitrarily designated, it is necessary to notify the image receiving device 3 of the designated arithmetic code data D5.
[0073]
Therefore, as shown in FIG. 15, the header generation unit 15 uses a 16-bit parameter called Rsiz defined as Table A-10 in the JPEG-200 system as a part of the packet header, for example, in the Rsiz. The arithmetic code data D5 designated to be scrambled is notified using the empty bits of the 16-bit information storage area JK1 that is unused except for all existing “0”.
[0074]
Here, when the arithmetic code data D5 is scrambled, the header generation unit 15 assigns “1” to the first bit in the empty bits of Rsiz, and conversely, scrambles the arithmetic code data D5. Is not applied, “0” is assigned to the first bit in the empty bits of Rsiz.
[0075]
Further, when “1” is assigned to the first bit in the free bit of Rsiz, the header generation unit 15 assigns wavelet transform level information to the subsequent 3 bits from the second bit to the fourth bit. Incidentally, since 3 bits are used for the wavelet transform level information, the 0 to 7 levels can be designated by the 3 bits.
[0076]
Further, the header generation unit 15 assigns a level to be scrambled to the fifth and subsequent bits with respect to the empty bits of Rsiz. Actually, for example, if the 5 bits from the 5th bit to the 9th bit are sequentially associated with the first level to the 5th level and “10010” is assigned to the 5 bits, the 1st level and the 4th level It is possible to notify that the second level, the third level, and the fifth level are not scrambled.
[0077]
In this way, the header generation unit 15 can appropriately notify the image receiving device 3 of information on the presence / absence of the scramble process as header data.
[0078]
Thereby, in the image receiving apparatus 3, the descrambling processing unit 23 can appropriately perform the descrambling process on the scrambled data D6 based on the information on the scramble process shown in the header data D7. As shown in FIG. 5, the image SGA can be visually recognized accurately.
[0079]
By the way, the image transmission apparatus 2 does not scramble the header data D7, but only scrambles the arithmetic code data D5.
[0080]
Therefore, if the image receiving device is configured similarly except for the image receiving device 3 and the descrambling processing unit 23 described above with reference to FIG. 3, when the packet stream D8 transmitted from the image transmitting device 2 is received, the packet Since encoding information necessary for decoding is stored in the header data D7 obtained based on the stream D8, together with arithmetic code data D5 (data not subjected to scramble processing at the time of transmission) based on the encoding information. The scrambled data D6 can be decoded as it is.
[0081]
In such an image receiving device, when image data is generated so as to decode the arithmetic code data D5 and the scrambled data D6, the scrambled data D6 cannot be properly decoded. As shown in FIGS. 17 to 19, an outline of an image pattern or the like can be visually recognized to some extent.
[0082]
Incidentally, as shown in FIG. 17, an image SGB designated to scramble only the first level and an image SGC designated to scramble only the fourth level as shown in FIG. As compared with the above, it can be seen that the image quality degradation is smaller when the scramble processing is applied to a small level.
[0083]
Further, if it is specified that the scramble process is applied to different levels at the same time as the image SGD specified to be scrambled at the fifth level and the sixth level shown in FIG. 19, it is scrambled to only one level. It can be seen that the strength of the scramble process can be selected more finely than the case where it is designated to perform the process.
[0084]
In this way, in the image providing system 1, even when the outline of the image is shown to some extent to a user who does not have the common key data, the image can be provided with fine control of the degree. .
[0085]
In the above configuration, in the image providing system 1, the image transmitting device 2 performs wavelet transform on the image data D1 of the image to be provided, and quantizes the obtained wavelet transform coefficient D2 to generate a quantized coefficient D3. Arithmetic code data D5 is generated such that the generated quantized coefficient D3 is sequentially encoded in the coding pass in units of code blocks.
[0086]
Then, the image transmitting apparatus 2 performs scrambling processing according to the designation on the arithmetic code data D5 in units of code blocks generated as described above, and the obtained scrambled data D6 is encoded to indicate the content of the encoding processing. The packet data is generated by adding the header data D7 storing the information, and the packet data is transmitted to the image receiving device 3.
[0087]
On the other hand, when the image receiving device 3 receives the packet data transmitted from the image transmitting device 2, the packet data is separated into header data D7 and scrambled data D6 and subjected to descrambling processing on the scrambled data. Based on the encoded information stored in the header data D7, the original arithmetic code data D5 is arithmetically decoded in the decoding pass to generate the quantized coefficient D11, and the quantized coefficient D11 is converted to the inverse quantum. Then, the wavelet transform coefficient D12 obtained is subjected to inverse wavelet transform to generate image data D13.
[0088]
Therefore, in this image providing system 1, since the image transmission device 2 encodes the image data D1 and then scrambles the obtained arithmetic code data D5, the amount of data after encoding the quantized coefficients is reduced. A significant increase can be prevented.
[0089]
In addition, in this image providing system 1, since the image transmission device 2 performs the scrambling process by dividing the arithmetic code data D5 into a predetermined data length by the block encryption method, the data amount of the arithmetic code data D5 is small. An increase after the scramble process can also be prevented.
[0090]
In the image providing system 1, even in the image receiving device 3 that receives the packet data transmitted from the image transmitting device 2, the descrambling process is reversible with the scramble process on the transmitting side, and the decoding is transmitted. Since it is reversible with the encoding on the side, as in the case of the image transmission apparatus 2, it is possible to prevent the data compression rate from being lowered until the received packet data is decoded through the descrambling process. Can do.
[0091]
According to the above configuration, in the image transmitting apparatus 2, the image data D1 is quantized after wavelet transform, and the obtained quantized coefficient D3 is arithmetically encoded in the encoding pass in units of code blocks to obtain arithmetic code data D5. , Scramble processing is performed on the generated code block unit arithmetic code data D5, header data D7 is added and transmitted to the image receiving device 3, and the image transmitting device 3 receives the received packet data. Is divided into header data D7 and scrambled data D6, descramble processing is performed on the scrambled data, and the obtained arithmetic code data D5 is obtained in the decoding pass based on the coding information stored in the header data D7. After performing arithmetic decoding with, inverse quantization of the obtained quantized coefficient D11, and inverse wavelet transform By generating the image data D13, the image transmission apparatus 2 prevents the amount of data from increasing after the quantization coefficient D3 is encoded and before being transmitted as packet data after being scrambled. In accordance with this, it is also possible to prevent an increase in the amount of data between reception of packet data and decoding by the image reception device 3 having a configuration corresponding to the image transmission device 2. Thus, although the image data is scrambled and transmitted, an image providing system that can prevent the data compression rate from decreasing can be realized.
[0092]
In the above-described embodiment, the scramble process is performed on a part or the whole of the image by arbitrarily specifying the individual arithmetic code data D5 in units of code blocks to be scrambled and the level and subband. However, the present invention is not limited to this, and as shown in FIGS. 20A and 20B, the top and bottom of each subband corresponds to the top and bottom of the image SGE. A part to be scrambled with respect to the image SGE may be specified, and the arithmetic code data D5 that is actually subjected to the scramble process corresponding to the specified part may be designated.
[0093]
In addition to this, the depth of the arithmetic code data D5, that is, the arithmetic code data D5 may be divided by a predetermined data length and specified to be subjected to the scramble processing. In this case, the arithmetic code data D5 may be specified. The scramble strength can be adjusted according to the data length for delimiting the data D5.
[0094]
In the above-described embodiment, the case has been described in which packet data is generated by adding arithmetic code data D5 in code block units or scrambled data D6 in code block units to header data D7. However, the present invention is not limited to this, and a hierarchical layer may be generated from a plurality of arithmetic code data D5, and packet data may be generated for each layer.
[0095]
Incidentally, in order to generate such packet data, for example, as shown in FIG. 21, before encoding the quantization coefficient D3 of the code block 0 to the code block 3 in the encoding unit 12, for example, the code block 0 to the code block For each block 3, each absolute value bit plane is associated with, for example, packet data 0 to packet data N so as to be hierarchically distributed, and as shown in FIG. In the absolute value bit plane associated with each code block 0 to code block 3 so as to be hierarchically distributed to packet data 0 to packet data N at the time when the quantization coefficient D3 of the code block 3 is encoded. Encoding results (that is, arithmetic codes) are arranged between hierarchical layers 0 to It is N, may be generated packet data to each of these layers 0 through Layer N.
[0096]
In FIG. 21, C indicates a cleanup path, S indicates a significance path, M indicates a refinement path, and 0 indicates that no path exists. The number of layers is set in advance so that the layer 0 to the layer N are associated with the packet data 0 to the packet data N in advance. In FIG. 22, each of layer 0 to layer N corresponds to packet data 0 to packet data N, and a block indicated by hatching for each of layer 0 to layer N represents the code amount of the arithmetic code by its height. In addition, Empty represents that there is no arithmetic code.
[0097]
Then, as shown in FIG. 23, a packet stream D20 is generated by adding header data D7 to layered arithmetic codes (Body), and a layer coefficient is set to λ from 0 (most significant) to 2,. In addition to Λ (lowest order), the resolution coefficient is 1 and is expressed from 0 (lowest resolution) to 1,..., L, (highest resolution), and the color component index is expressed as 1, 2,. For example, when there is one color component index C, packet data belonging to the same layer (the same value of λ) is collected in order from the corresponding spatial resolution from the lowest to the higher, and the collected packets SNR (Signal to Noise Ratio) progressive can be realized by arranging the data in order from the lowest layer coefficient λ to the higher one (λ from 1 to Λ).
[0098]
Further, as shown in FIG. 24, when the packet stream D21 is generated by adding the header data D7 to the layered arithmetic code (Body) and the color component index C is one, the same spatial resolution is obtained. Corresponding packet data are arranged in order from the lowest layer coefficient λ (λ is from 1 to Λ), and the packet data in which these are arranged is arranged from the lowest spatial resolution to the higher (l from 0 to L) If they are arranged in order, resolution progressive can be realized.
[0099]
When progressive is realized in this way, as shown in FIG. 25, the header generation unit 15 of the image transmission apparatus 2 is, for example, an 8-bit parameter called “Scode” defined as Table A-10 in the JPEG-2000 system. Thus, if the information storage area JK2 in which the upper 4 bits of the 8 bits are vacant bits (Reserved) is used, the layer number and the like can be notified to the image receiving apparatus.
[0100]
Further, in a layered packet stream, for example, a scramble process may be performed only on an arithmetic code in packet data existing in a specific layer, such as one having all resolutions with a layer coefficient λ of 0. That is, in SNR progressive, the higher the layer, the greater the effect on the image quality. Therefore, if the scrambling process is performed on the arithmetic code in the packet data existing in the upper layer, the scrambling process is stronger than that in the lower layer. Can be applied.
[0101]
Furthermore, instead of designating only one layer to be scrambled, for example, a specific number of layers counted from the highest layer or a specific number of layers counted from the lowest layer You can also.
[0102]
Further, in the above-described embodiment, the case where the quantization coefficient D3 is encoded in units of code blocks each having 64 samples in length and width has been described, but the present invention is not limited to this, and in EBCOT, code blocks Since the number of vertical and horizontal samples is specified as a power of 2 from 4 to 256, code blocks of various sizes such as code blocks of 32 samples each in length and width and 128 samples in length and 32 samples in width The quantization coefficient D3 may be encoded, or the code block size may be arbitrarily changed.
[0103]
Further, in the above-described embodiment, the case where the present invention is applied to the image transmitting device 2 and the image receiving device 3 of the image providing system 1 described above with reference to FIGS. 1 to 25 has been described. However, the present invention is not limited to this, and any device that encodes and / or decodes an image data recording / reproducing apparatus or video camera having both configurations of encoding and decoding images, and receiving images are provided. The present invention can be widely applied to image encoding devices and image decoding devices having various configurations such as portable telephones, personal computers, and PDAs (Personal Digital Assistance).
[0104]
Furthermore, in the above-described embodiment, the case where the wavelet transform unit 10 is applied as filter coefficient generation means for generating filter coefficients by filtering image data using a low-pass filter and a high-pass filter will be described. However, the present invention is not limited to this, and filter coefficient generation means having various other configurations can be widely applied as long as the filter coefficient can be generated by filtering the image data using a low-pass filter and a high-pass filter. be able to.
[0105]
Furthermore, in the above-described embodiment, a case has been described in which the encoding unit 12 is applied as an encoding unit that generates arithmetic code data by encoding quantization coefficients using a predetermined encoding method. However, the present invention is not limited to this, and it is possible to widely apply encoding means having other various configurations as long as it can generate the arithmetic code data by encoding the quantization coefficient by a predetermined encoding method. Can do.
[0106]
Furthermore, in the above-described embodiment, based on the encoding information stored in the header data, the arithmetic code data restored by the descrambling processing means is decoded by a predetermined decoding method to be quantized. Although the case where the decoding unit 25 is applied as the decoding means for generating the encoding coefficient has been described, the present invention is not limited to this, and the descrambling is performed based on the encoding information stored in the header data. If the arithmetic code data restored by the processing means can be decoded by a predetermined decoding method to generate a quantized coefficient, decoding means having various other configurations can be widely applied.
[0107]
Furthermore, in the above-described embodiment, when the wavelet inverse transform unit 27 is applied as image data generation means for generating image data by filtering the filter coefficients using a low-pass filter and a high-pass filter. However, the present invention is not limited to this, and image data generating means having various configurations can be used as long as filter data can be filtered using a low-pass filter and a high-pass filter to generate image data. Can be widely applied.
[0108]
【The invention's effect】
As described above, according to the present invention, wavelet transform for recursively filtering low frequency components is performed on image data to generate wavelet coefficients in subband units as filter coefficients, and the generated wavelet coefficients are quantized. To generate a quantized coefficient, encode the quantized coefficient with a predetermined encoding method to generate arithmetic code data, and then scramble the specified part of the arithmetic code data in an arbitrary subband unit And generating predetermined coding information relating to the coding process based on the arithmetic code data, and adding the scrambled arithmetic code data to the header data generated by storing the generated coding information. Packet data is generated by encoding the quantization coefficient, and then the encoding process is performed. As a result of performing scramble processing on the specified parts of arbitrary subband units of the arithmetic code data obtained in this way, image quality degradation can be reduced for images based on image data, and quantization coefficients are encoded Sometimes it is possible to prevent the amount of data from increasing, thus preventing a reduction in the compression rate of the image data.
[0109]
According to the present invention, the wavelet transform for recursively filtering the low frequency component is applied to the image data, and the subband unit wavelet coefficients generated as filter coefficients are quantized and then subjected to predetermined encoding. Arithmetic code data obtained by performing scramble processing on a specified part in an arbitrary subband unit is added to header data storing encoding information when arithmetic code data is generated by the encoding method. The header data and the scrambled arithmetic code data are separated from the generated packet data, the scrambled arithmetic code data is descrambled to return to the original arithmetic code data, and the header Based on the encoding information stored in the data, the original arithmetic code data is decoded to the predetermined After generating the quantized coefficient by decoding with the equation, and generating the filter coefficient by dequantizing the generated quantized coefficient, the filter coefficient is filtered using the low pass filter and the high pass filter By generating image data, if this configuration is used, the image data encoding side will have a reversible configuration, and after encoding the quantized coefficients, the code As the scramble process is applied to the specified part of the arbitrary subband unit of the arithmetic code data obtained by the conversion process, image quality degradation can be reduced for the image based on the image data, and the quantization coefficient is encoded. It is possible to prevent the amount of data from increasing when the image data is converted, and thus to prevent a reduction in the compression rate of the image data.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an embodiment of a configuration of an image providing system according to the present invention.
FIG. 2 is a block diagram illustrating a circuit configuration of the image transmission apparatus.
FIG. 3 is a block diagram illustrating a circuit configuration of the image receiving apparatus.
FIG. 4 is a schematic diagram for explaining wavelet transform;
FIG. 5 is a schematic diagram showing subbands by first-level wavelet transform.
FIG. 6 is a schematic diagram showing subbands up to a second level.
FIG. 7 is a schematic diagram illustrating subbands obtained by wavelet transform up to a second level.
FIG. 8 is a schematic diagram showing subbands up to third and fourth levels.
FIG. 9 is a schematic diagram for explaining quantization in a quantizer;
FIG. 10 is a schematic diagram for describing encoding in an encoding unit.
FIG. 11 is a schematic diagram for explaining generation of a bit plane in an encoding unit.
FIG. 12 is a schematic diagram for explaining scanning of an absolute value bit plane in an encoding unit;
FIG. 13 is a schematic diagram for describing encoding of each absolute value bit plane for one code block;
FIG. 14 is a schematic diagram for explaining the designation of arithmetic code data to be scrambled.
FIG. 15 is a schematic diagram showing a configuration of Rsic used as part of packet data.
FIG. 16 is a schematic diagram for explaining an image obtained by performing descrambling processing;
FIG. 17 is a schematic diagram for explaining an image in which a first level is scrambled.
FIG. 18 is a schematic diagram for explaining an image in which a fourth level is scrambled.
FIG. 19 is a schematic diagram for explaining an image obtained by performing scrambling on the fifth level and the sixth level.
FIG. 20 is a schematic diagram for explanation when a scramble process is performed on an arbitrary part of an image according to another embodiment;
FIG. 21 is a schematic diagram for explanation when an absolute value bit plane in each code block is associated with packet data in advance.
FIG. 22 is a schematic diagram for explaining packet data layering according to another embodiment;
FIG. 23 is a schematic diagram for explanation when realizing SNR progressive by layers;
FIG. 24 is a schematic diagram for explaining the case of realizing resolution progressive by layers.
FIG. 25 is a schematic diagram for explaining a Scod used as a part of packet data.
[Explanation of symbols]
DESCRIPTION OF SYMBOLS 1 ... Image provision system, 2 ... Image transmission apparatus, 3 ... Image reception apparatus, 10 ... Wavelet transformation part, 11 ... Quantization part, 12 ... Encoding part, 13 ... Rate control part, 14 ...... Arithmetic code extraction unit, 15 ... Header generation unit, 16 ... Scramble processing unit, 17 ... Packet generation unit, 21 ... Packet decoding unit, 22 ... Header decoding unit, 23 ... Descramble processing unit, 24 …… Arithmetic code decoding unit, 25 …… Decoding unit, 26 …… Dequantization unit, 27 …… Wavelet inverse transformation unit, HH1, HH2, HH3, HH4 …… Highest band subband, HL1, HL2, HL3 , HL4... High and low frequency subbands, LH1, LH2, LH3, LH4... Low and high frequency subbands, LL1, LL2, LL3, and LL4... ...... wavelet transform coefficients, D3, D11 ...... quantized coefficients, D4, D5 ...... arithmetic code data, D6 ...... scrambled data, D7 ...... header data, D8 ...... packet stream.

Claims (6)

A wavelet coefficient generating means for generating a wavelet coefficient of the sub-band unit as the filter coefficients by performing wavelet transform to filter recursively low-frequency components for the image data,
Quantization means for quantizing the wavelet coefficients to generate quantization coefficients;
Encoding means for generating arithmetic code data by encoding the quantized coefficient according to a predetermined encoding method;
A scramble processing means for performing a scramble process on a specified part of an arbitrary subband unit of the arithmetic code data;
Header data generating means for generating predetermined encoding information related to the encoding process based on the arithmetic code data, and generating header data storing the generated encoding information;
An image encoding apparatus comprising: packet data generation means for generating packet data by adding the arithmetic code data subjected to the scramble processing to the header data. Separation means for separating the header data and the scrambled arithmetic code data from the packet data;
Descrambling processing means for performing descrambling on the arithmetic code data subjected to the scramble processing and returning the arithmetic code data to the original arithmetic code data;
Based on the coding information stored in the header data, and decoding means for generating the quantized coefficients of the arithmetic code data returned to the original processing decoded in a predetermined decoding method,
Inverse quantization means for inversely quantizing the quantization coefficient to generate the filter coefficient;
The image coding apparatus according to claim 1, further comprising: an image data generation unit configured to filter the filter coefficient using a low- pass filter and a high-pass filter to generate the image data. A wavelet coefficient generation step of generating a wavelet coefficient of the sub-band unit as the filter coefficients by performing recursive filtering to wavelet transform low-frequency components for the image data,
A quantization step for quantizing the wavelet coefficient to generate a quantization coefficient;
An encoding step for encoding the quantized coefficient by a predetermined encoding method to generate arithmetic code data;
A scramble processing step of performing a scramble process on a specified part of an arbitrary subband unit of the arithmetic code data;
A header data generation step of generating predetermined encoding information related to the encoding process based on the arithmetic code data and generating header data storing the generated encoding information;
And a packet data generation step of generating packet data by adding the arithmetic code data subjected to the scramble processing to the header data. A separation step of separating the header data and the scrambled arithmetic code data from the packet data;
A descrambling process step of performing descrambling on the arithmetic code data subjected to the scrambling process and returning the arithmetic code data to the original arithmetic code data;
Based on the coding information stored in the header data, a decoding step of generating the quantized coefficients of the arithmetic code data returned to the original processing decoded in a predetermined decoding method,
An inverse quantization step of dequantizing the quantization coefficient to generate the filter coefficient;
The image encoding method according to claim 3 , further comprising: an image data generation step of generating the image data by filtering the filter coefficient using a low- pass filter and a high-pass filter. The wavelet transform that recursively filters the low-frequency components of the image data is subjected to quantization, and the wavelet coefficients in subband units generated as filter coefficients are quantized and then encoded using a predetermined encoding method to perform arithmetic From the packet data generated by adding the above arithmetic code data subjected to the scramble processing to the specified portion of any subband unit to the header data storing the encoding information when the code data is generated, Separation means for separating the header data and the scrambled arithmetic code data;
Descrambling processing means for performing descrambling on the arithmetic code data subjected to the scramble processing and returning the arithmetic code data to the original arithmetic code data;
Decoding that generates a quantized coefficient by decoding the arithmetic code data restored by the descrambling processing means using a predetermined decoding method based on the encoding information stored in the header data And
Inverse quantization means for inversely quantizing the quantization coefficient to generate the filter coefficient;
An image decoding apparatus comprising: image data generating means for generating the image data by filtering the filter coefficient using a low- pass filter and a high-pass filter. The wavelet transform that recursively filters the low-frequency components of the image data is subjected to quantization, and the wavelet coefficients in subband units generated as filter coefficients are quantized and then encoded using a predetermined encoding method to perform arithmetic From the packet data generated by adding the above arithmetic code data subjected to the scramble processing to the specified portion of any subband unit to the header data storing the encoding information when the code data is generated, A separation step of separating the header data and the scrambled arithmetic code data;
A descrambling process step of performing descrambling on the arithmetic code data subjected to the scrambling process and returning the arithmetic code data to the original arithmetic code data;
A decoding step of generating a quantized coefficient by decoding the arithmetic code data returned to the original by a predetermined decoding method based on the encoding information stored in the header data;
An inverse quantization step of dequantizing the quantization coefficient to generate the filter coefficient;
An image decoding method comprising: an image data generation step of generating the image data by filtering the filter coefficient using a low- pass filter and a high-pass filter.

Images