TWI373959B - Wavelet codec with a function of adjustable image quality - Google Patents

Description

1373959 IX. Description of the invention: [Technical field to which the invention pertains] The arrangement is particularly concerned with the device. The present invention relates to an image compression processing apparatus for adjusting image quality of wavelet transform coding processing. [Prior Art]

In today's network environment, the application of multimedia has played an important role. 'The requirements for image compression encoders are becoming more and more stringent'. In addition to the requirement for high-quality images, the original image quality will still be More advanced steps require image editing to support a variety of features, such as control of restoration quality, control of reduced resolution, search capabilities with regions of interest (r〇i, Region Of lnterest), etc. In response to these requirements, many different image compression coding standards have emerged in recent years. Due to some attractive features, wavelet transform has proven to be very useful for encoding video and video. Therefore, the image encoders proposed in recent years are mostly based on wavelet transform. Well-known coding standards, such as EZW (Embedded Zerotree Wavelets), use a tree structure, SPIHT (Set Partitioned in Hierarchical Trees) uses a serial structure, and SPECK (Set Partitioned Embedded Block Coder) uses zero blocks and strings simultaneously. Embedded Wavelet Image Coding Based on Zero-Blocks and Array (EZBA) and Improved Zero-Block and Array Embedded Wavelet Coding (Improved Embedded Wavelet Image Coding)

Based on Zero-Blocks and Array, I-EZBA) uses a zero block and array structure. —6 — 1373959

The coding standards mentioned above only support distortion adjustment, not resolution adjustment and (10) selection ability. In order to solve this problem, Gu.O〇e proposed the S_SPECK coding standard, which is extended by spEcK to enable the editing to support three characteristics of distortion, adjustable resolution and ROI selection. Does not add too much computational complexity. However, the coded bits generated by the S-SPECK encoded compression core are relatively scattered, and the codes compiled by the coefficients of the same unit code (codeunit 'or sul>band) are not necessarily placed together. In order to obtain the resolution adjustability, the numerator bits το belonging to the same unit code must be placed at _', so it is necessary for the Xiangji ship to collect these coded bits first, and then fill in the character f layer. Therefore, a large amount of memory space is required at the stage of forming the u layer, and a lot of extra recording bits are used in the encoded stream. SUMMARY OF THE INVENTION The main object of the present invention is to provide a wavelet transform coding processing apparatus capable of adjusting image quality, which has a projection degree of distortion, a light resolution, and a selection of at least a region of interest, so as to greatly reduce the encoding process. The amount of A memory space required for the formation phase of the pin layer, and the difficulty of making a small number of extra bits in the 'flat code stream, make the invention have the effect of reducing cost and computational complexity. The purpose of this month is to provide a wavelet transform coding processing device that can adjust the image quality, which can be used to compress and compile the condensed coding operation and the layer control (n) step. The invention has the effect of improving coding efficiency. 7~ The image-converted wavelet transform coding processing apparatus according to the present invention comprises a discrete wavelet transform unit, a wavelet coefficient grouping unit, a compression coding unit and a quality layer controller. The discrete wavelet transform unit performs two-dimensional wavelet transform on an input image, and generates a wavelet coefficient matrix corresponding to the input image size; the wavelet coefficient grouping unit is configured to receive the wavelet coefficient matrix, and the coefficients in the wavelet coefficient matrix Performing coefficient recombination corresponding to a plurality of ROIs; the compression coding unit receives the recombined coefficient 'and encodes using a zero block coding technique; the quality layer controller compresses the encoded data according to several The preset bit budget and several resolution levels are filled in several quality layer formats. The above and other objects, features, and advantages of the present invention will become more apparent from the <RTIgt; The adjustable image area (4) of the preferred embodiment of the present invention, the code conversion processing device 1 is Scalable ROI Image Wavelet S, and _ _ coder , :; wave to rr processing device single; wave rotation; (four) ship base transposition 1 discrete small conversion, generate ^ ^ compressed image to perform two-dimensional wavelet "original shirt image size of the wavelet coefficient matrix; 8 - the coefficient grouping unit 12 is configured to receive the wavelet coefficient matrix 're-arrangement and set ' through a coefficient grouping sub-unit 121 and a ROI block grouping sub-unit 122 to perform coefficient recombination of the wavelet coefficient matrix, so that the encoding The system 1 obtains the retrievability of the Region of Interest (ROI) and the adjustability of the resolution, wherein the coefficient sub-unit 121 is used to

The coefficients of the wave coefficient matrix are grouped to form a plurality of r0I blocks, and the R〇I block grouping subunit 122 divides the plurality of R〇1 blocks into a plurality of unit code units (Codeunit); the compression coding unit 13 can receive the recombined coefficient 'and encode using I-EZBA coding technology; the quality layer controller 14 &quot;T according to the user's needs, after setting different bit budgets and resolution levels, 'compress coding The subsequent bits are filled in the format defined by the S_EZBA's editing system 1 to form a bit stream that meets the user's requirements, and a decoding device is used to select the image quality and resolution to be restored when the image is restored. degree. The above I-EZBA coding uses the method of directly extracting the feature values to complete the feature value extraction processing of the image. Compared with the conventional EZBA code, the coding mode of the coefficient correspondence map needs to be established, and the ^EZBA code does not require the establishment of the correspondence map of the important coefficients. The purpose of reducing the amount of memory used can be achieved. In order to obtain the ROI search capability, all wavelet coefficients of the image to be compressed can be reconstructed into a spatial steering tree, and the spatial steering tree is independently encoded. As shown in Fig. 2, it is revealed that the spatially oriented tree exhibits a parent-child coefficient relationship in different sub-bands LL3, LH3, HL3, HH3, LH2, HL2, HH2, LH1, HL1, and HH1, and all coefficients are organized into a tree structure. And the root of the tree is in the lowest frequency band LL3; Figure 2-9 — 1373959

The coefficients in the two frequency bands are spatially corresponding to a certain one of the original images. The parent-child coefficients of the spatially-oriented tree correspond to the same position of the original image towel in the human band towel. In the original image, the square domain can be reconstructed by the coefficients of the individual spatial steering tree. If the large J of the lowest frequency band LL3 in Fig. 2 is 4 4 or 16 coefficients, there are 16 spatial orientations. The tree can be used to reconstruct 16 corresponding ROI images. The original image can be decomposed into different sub-bands by wavelet transform, and the K-segment decomposition can generate 3K+1 sub-bands and κ+1 layer resolution level-month reference. Figure 3 does not reveal a relationship between a sub-band structure undergoing three-stage (κ=3) wavelet transform and a four-layer resolution level RQ, R1, Μ and ' 'the lowest resolution layer' level R〇 The lowest frequency sub-band LL3 is composed; the second low resolution level R1贱 additionally includes three high-frequency sub-bands HL3, LH3 and flip, and the next resolution level is increased in a similar manner. Frequency band. When a certain solution is needed In the case of image degree, as long as it is not included in this resolution (10), it is necessary to perform the reduction in the sub-band of the remaining ^. For example, when the resolution is required, the three sub-bands HU, Cong and Lm is excluded, and the coefficients in the three sub-bands HU, HH1, and LH1 are not encoded. Therefore, the adjustability of the resolution is obtained, and the sub-band (10) money coefficient is divided into different sub-sets according to the resolution level, and The wavelet coefficient grouping unit of the preferred embodiment of the present invention is described in detail. In order to achieve simultaneous ROI retrieval capability and resolution adjustability, the wavelet must be first -10 - 1373959. The Wei Wei organization becomes Wei Gelang, and the coefficients in the space-oriented tree are grouped according to the resolution, , and the coefficients of the same resolution level to form a unit code as shown in Fig. 4 (Codeunit) CO, Cl, C2, turn + no open tongue's built-in ability can be obtained by the number of spaces parsed by the wavelet coefficient matrix, and then the space guide tree distinguishes the several Unit ° I, C2, so Each of the R〇l obtains the adjustability of the resolution, wherein the early generations of the mothers CO, a, and C2 are the basic units of the compression coding, as shown in the figure, because the minimum unit of the zero block pressure is to be performed. At least one block of 2χ2 is required, so (3) at least 2x2 blocks, so even if the wavelet transform in the 3 stage can be the area = 4: solution: the sub-band of the degree level, but in order to yield the small block of the zero block , so it will be in the first layer resolution R 〇 [ = ], and the three coefficients of the second layer resolution together constitute the single C0 ', therefore, only three resolution levels are separated in the space-oriented shot, The early code co is composed of the coefficients in the sub-bands lu and ^, LH3, HH3 in the space-oriented tree. The unit code α is formed by the sub-bands HL2, LH2, (4) in the space=, the unit The code q consists of the sub-bands, _, Hm in the two leading trees. After the residual weight _ of the wavelet residual matrix is over, the subsequent coded encoding operation can be performed by using the code. Referring to FIG. 5, in the compression coding unit 13, the compression coding operation of the coefficients in the (10) block of the present invention utilizes a 2-word coding order, that is, the unit code is from low resolution to high. The resolution sequence is encoded, and the rounding of its encoding bit 〆-11-1373959 is as shown in Fig. 6, from the high bit plane to the low bit plane, the bit reading is simple. The code c: 'C2' It is easy to distinguish between == resolution layer and (4) code although it is convenient to solve the problem. _Tree_Cheerba performs: Code operation, so compared with the comparison result of the conventional coding technique _ coefficient and gate pinch value, the complexity of the cake low-margin operation is saved, and the use of a large amount of memory is saved. .

Referring to FIG. 7a and FIG. 7b, the quality layer format 2 defined by the wavelet transform coding processing apparatus 1 of the present invention is disclosed. After each R〇i block is compression-encoded, the coded bits that can be rotated are displayed. The meta is assigned to this format 2. The quality layer format 2 is provided with a plurality of quality layers, and the preferred embodiment of the present invention is as an example. The quality layer format 2 package order--the first quality belongs to Q〇, one

a first quality layer Q1, a second quality layer Q2, a fourth quality layer Q3, and a fifth quality layer Q4, which are used to enable the selection of the quality layers Q0 to Q4 when the ROI block is reconstructed. Decoding is performed to select the reconstruction of the image quality of the ROI block. As shown in FIG. 8, the plurality of quality layers Q0 to Q4 can be used as a boundary by a bit rate. For example, when a quality of 2525bpp (bit per pixel) is required, the first one can be selected. The quality layer Q0 and the second quality layer Q1 are decoded. In the first quality layer Q0 of the quality layer format 2, an additional extra bit and an interleaved bit 23 are included in the first quality layer Q0. Further, a first field 21 and a second field 22 are further disposed in the extra bit of the quality layer Q, the first booth is used to record the length of the quality layer. The second paste 22 is used to record the first bit. 12 — 1373959 The number of resolution levels that can be provided by the ROI block. In addition, as shown in FIG. 7b, the extra bits including all the quality layers Q2 to Q4 after the second quality layer Q1 are only provided with the first Field 21 and the interleaved coded bit 23. The interleaved coding bit 23 can be used to directly fill the output of the compression coding unit 13 until the length of the quality layer reaches the length recorded by the first barrier 21, and the bit of the next quality layer is performed. When allocating, the first interceptor 21 is also output first, and then the interleaved coded bit 23 of the previous quality layer is continued, and the subsequent coded bits are filled into the interleaved coded bit 23 of the quality layer until the bit The budget has arrived. Referring to FIG. 9, a circuit block diagram of the compression coding unit 13 of the wavelet transform coding processing apparatus 1 is disclosed, wherein the circuit block diagram is suitable for processing R〇i blocks of 8*8 pixel size. The circuit block diagram of the compression coding unit 13 includes a unit code reader 131, a coefficient processing module 132, a 16-bit temporary register 133, a symbol memory element 134, a decision encoder 135, a coefficient/ The symbol synthesis module 136, a symbol flag 137, and a threshold generator 丨38. The unit code reader 131 reads the coefficient of the ROI block stored in an external memory; the unit code reader 131 sends the coefficient to the coefficient processing module 132, and the coefficient processing module 132 sets the coefficient. The symbol bit is retrieved and stored in the symbol memory component 134, and the coefficient processing module 132 extracts the binary bit plane value and inputs the 16-bit scratchpad 133, and then the decision encoder 35 reading the binary value of the coefficient to judge the important coefficient or block and output the coding bit of the first stage, and then the coefficient/symbol synthesis module 13 6 will encode the bit and the corresponding symbol - 13 - 1373959 曰De Lingkou temporarily thinks about the coding bit, that is, the result of the I-coupling code. After the U-layer controller 14 takes the required bit budget and the resolution level, the user-defined bit is filled into each quality layer, and Immediately read the coded position to the job budget, #Quality layer controller Μ, (4) ― signal to inform the early generation to correct (four) 丨, - do (four) code coefficient.

As described above, the wavelet transform coding processing apparatus i of the present invention can reduce the coding and the output of the quality layer when assigning the == quality layer::=, does not need to wait until the bit budget of the quality layer arrives. , input, and do not need to use the memory space to collect the coding bits belonging to the same 2 code position to reduce the memory 2, so that the invention has the cost reduction, the computational complexity and the extraction code The present invention has been described with reference to the preferred embodiments of the present invention, and is not intended to limit the invention, and various modifications may be made to the above embodiments without departing from the spirit and scope of the invention. The modification is still the technical material protected by the present invention, and therefore the invention is defined by the scope of the patent application. 1373959 [Simple Description of the Drawing] FIG. 1 : Preferred Embodiment of the Invention Schematic diagram of the encoding process of the wavelet transform encoding processing device capable of adjusting image quality. FIG. 2 is a wavelet transform encoding processing device capable of adjusting image quality according to a preferred embodiment of the present invention. Schematic diagram of the interleaving tree. Fig. 3 is a schematic diagram showing the resolution level of the wavelet transform encoding processing device capable of adjusting image quality according to a preferred embodiment of the present invention. Fig. 4 is a diagram showing wavelet transform of an adjustable image quality according to a preferred embodiment of the present invention. Schematic diagram of the composition of each unit code of the encoding processing device. Fig. 5 is a schematic diagram of the zigzag encoding of the wavelet transform encoding processing device capable of adjusting image quality according to a preferred embodiment of the present invention. Fig. 6 is a view showing a preferred embodiment of the present invention. FIG. 7a is a schematic diagram of a quality layer format of a wavelet transform encoding processing device capable of adjusting image quality according to a preferred embodiment of the present invention. FIG. 7b is a schematic diagram of a quality layer format of a wavelet transform encoding processing device capable of adjusting image quality. A quality layer format diagram of a wavelet transform coding processing apparatus capable of adjusting image quality according to a preferred embodiment of the present invention. FIG. 8 is a diagram showing a wavelet transform coding processing apparatus capable of adjusting image quality according to a preferred embodiment of the present invention. A schematic diagram of several quality layer boundaries. Figure 9: Small image quality of the preferred embodiment of the present invention Circuit block diagram of the compression coding unit of the conversion coding processing device. [Description of main component symbols] 1 Wavelet transform coding processing apparatus 11 Discrete wavelet transform unit-15 - 1373959 12 Wavelet coefficient grouping unit 122 ROI block grouping subunit 131 unit code reading Picker 133 16-bit temporary register 135 decision encoder 137 symbol flag 14 quality layer controller 21 first field 23 interlaced coded bit 121 coefficient grouping sub-unit 13 compression coding early. element 132 coefficient processing module 134 symbol Memory Element 136 Coefficient/Symbol Synthesis Module 138 Threshold Generator 2 Encoding Format 22 Second Field

Claims (1)

Apply for a full-time 101 丨丨. The wavelet transform coding processing device of the adjustable image quality is modified in the following year, which comprises a political wavelet transform unit, which performs two-dimensional wavelet transform on an input image and generates a wavelet coefficient matrix for the input image size; The wave coefficient group TG ' is used to connect (4) the money matrix, and the coefficients of the coefficient coefficient moment (4) are recombined by the coefficients of the number of positions and the interest region r; the coding unit receives the recombined coefficient and uses block coding. The technology encodes; and the Z quality layer controller, the data after the encoding is based on a plurality of preset 兀 pre-existing and a plurality of resolution levels to fill the number of the product f-layer format; Lu, the coding unit of the surface includes - the unit code Reader, a coefficient LT, a 16-bit scratchpad, a symbol memory component, a decision coefficient/symbol synthesis module, a symbol flag, and a gate=storage unit. The unit code reader is used to read The (10) block will also be turned on, and the material and the number of processing modules will be stored in the memory element module to touch the H-element plane value, and the decision coefficient 11 will be read by the decision code 11 and the output order will be output. The segment of the segment, the flat code bit 兀, the symbol/symbol human symbol bit is merged, and the coded bit and the corresponding coded bit are sent to each of the quality: coded bits, and the negative layer &amp; In the state, to complete the output of the coded bits of a unit code. 2. The wavelet transform coding processing apparatus of the adjustable image quality according to the application scope of the patent application, wherein the wavelet coefficient grouping unit comprises: a ^^ group sub-it and a threat block sub-group ^, the coefficient The sub-group is used to group the coefficients of the wavelet coefficients _ to form a plurality of 1-blocks. The ROI block sub-group divides the plurality of biliary blocks into a plurality of unit codes. Li Fan (4) The image-quality wavelet transform and code processing device described in item 1 wherein the wavelet coefficients are recombined—the spatial steering tree skew=inter-steering tree is used to reconstruct a corresponding R0I block and independently performed in the guiding tree. The code 'the spatially oriented tree exhibits a parent-child coefficient relationship in different sub-bands. 4, = specializes in the wavelet transform of the adjustable image quality described in item 1, and the flat code processing is performed, and the K p ratio ^^^ in the basin produces 3K+I:f wavelet transform for decomposing the frequency band and K+ 1 layer resolution level. 1 Said that the fine image of the adjustable image quality mentioned in the second paragraph of the patent scope is low-resolution resolution sequence t code early reading the unit code by 6. According to the application of the patent scope, the conversion code processing! Jin, 1 卩 像 像 口 负 第 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - [When the block is rebuilt, the image quality of the wavelet is transferred to the product according to item 6 of the application. The quality layer format includes - the first layer and the second quality layer. On the 12th, the replacement page-changing code processing device is modified, wherein the five quality layers are demarcated by a bit rate. 8. The image-converted wavelet transform coding processing device according to claim 6 of the patent application scope, wherein each The quality layer includes an extra bit and an interlaced bit, the extra bit of the first quality layer is provided with a first field and a second block, and the second quality layer, the third quality layer, and the fourth Each of the quality layer and the fifth quality layer respectively has a first block, the first block is used to record the length of the quality layer, and the second block is used to record the resolution that the ROI block can provide. The number of levels, the interleaved coded bits are used by the compression coding unit to directly fill the output into the portion. 9. The image-converted wavelet transform coding processing device according to claim 1 of the patent application scope, wherein the quality layer The controller is completing one After the output coded bits of the code bits, sent a signal to notify the unit code reader to read the next coefficient.