Coding apparatus for digital image signals.

0001This invention relates to coding apparatus for digital image signals which apparatus can, for example, be used in a digital video tape recorder (VTR).

0002For coding a video signal, methods are known in which a mean number of bits (bit length) per pixel or a sampling frequency is reduced in order to narrow a transmission band. We have proposed coding apparatus in which a dynamic range which is specified by the maximum and minimum values of a plurality of pixels included in a two or three-dimensional block is obtained, and encoding adapted to the dynamic range is executed (Japanese patent specifications JP-A-61-144989 and JP-A-62-92620). Further, as disclosed in Japanese patent specification JP-A-62-128621, there has been proposed a variable length coding method in which the bit length is changed in accordance with the dynamic range, so that the maximum distortion occurring upon quantization is set to be constant.

0003According to the above coding method adapted to the dynamic range (ADRC), the dynamic range DR (difference between the maximum value MAX and the minimum value MIN) is calculated for each two-dimensional block comprising, for instance, 64 pixels (= 8 lines x 8 pixels). The minimum level (minimum value) in the block is eliminated from the input pixel data. The pixel data after elimination of the minimum value are converted into a representative level. The quantization includes dividing the dynamic range DR which has been detected into four level ranges corresponding to a bit length such as two bits smaller than the original quantization bit length, for detecting the level range to which each pixel datum in the block belongs, and for generating a code signal indicative of the level range.

0004We have also proposed a coding apparatus in which a three-dimensional ADRC and a frame dropping process are combined in order to raise the compression ratio further. In this apparatus, motion between two areas forming a three-dimensional block is detected, three-dimensional ADRC is executed in a block having motion, and in a stationary block, the transmission of one of the areas is omitted and two-dimensional ADRC is executed with respect to a block comprising a mean value of two areas.

0005In the above apparatus in which three-dimensional ADRC and frame dropping are combined, memory capacity of two frame is needed for a block segmentation process to form a three-dimensional block. When a buffering process is applied to the above apparatus, a memory of two frames is further necessary to delay input data until the number of bits to be assigned has been decided by the buffering process. A memory capacity of two frames is also needed for a block desegmentation of a decoding apparatus. The total memory capacity required is therefore large.

0006According to the present invention there is provided coding apparatus for a digital image signal, the apparatus comprising: means to which a digital image signal which has been converted into a signal having a two-dimensional block structure is supplied, and which is such that between first and second blocks located at spatially the same position and respectively belonging to two sequential frames, said first block is always transmitted and transmission/non-transmission of the second block is adaptively executed; control means for controlling the transmission/non-transmission of the second block in accordance with the absolute value of a difference between values of pixel data at the same position in said first and second blocks; and coding means to which the pixel data of the first block and the pixel data of the second block which has been decided to be transmitted are supplied, and which compression encodes said pixel data on a block unit basis.

0007According to the present invention there is also provided a coding apparatus for a digital image signal, the apparatus comprising: sub-sampling means to which a digital image signal which has been converted into a signal having a two-dimensional block structure is supplied, and in which a sub-sampling phase is inverted every two sequential frames;

0008means to which an output of said sub-sampling circuit is supplied and which is such that between first and second blocks which are located at spatially the same position and respectively belong to two sequential frames, the first block is always transmitted and a transmission/non-transmission of the second block is adaptively executed; control means for controlling the transmission/non-transmission in accordance with an absolute value of a difference between values of pixel data at the same position between said first and second blocks; and coding means to which the pixel data of the first block and the pixel data of the second block which has been decided to be transmitted are supplied and which compression encodes said pixel data on a block unit basis.

0009The invention will now be described by way of example with reference to the accompanying drawings, throughout which like parts are referred to by like references, and in which: <ul id="ul0001" list-style="none"><li>Figure 1 is a block diagram of an embodiment of the invention;</li><li>Figure 2 is a schematic diagram for explaining sub-sampling;</li><li>Figure 3 is a block diagram of an example of a buffering circuit;</li><li>Figure 4 is a schematic diagram of an example of a threshold value table;</li><li>Figure 5 is a flowchart for explaining the operation of the buffering circuit;</li><li>Figure 6 is a schematic diagram showing a frequency distribution table and an accumulative frequency distribution table which are formed by the buffering circuit;</li><li>Figure 7 is a flowchart for explaining the operation of an embodiment; and</li><li>Figure 8 is a schematic diagram for explaining the embodiment.</li></ul>

0010Referring to Figure 1, in the embodiment a digital video signal is supplied via an input terminal 1 to a block segmentation circuit 2 which converts-the video signal from a raster scan into a block signal. That is, a picture plane of one frame is segmented into two-dimensional blocks of, for example, (4 lines x 8 pixels). Unlike three-dimensional blocks, when two-dimensional blocks are formed, the memory capacity of the block segmentation circuit 2 can be reduced.

0011An output signal from the block segmentation circuit 2 is supplied to a sub-sampling circuit 3 which halves the number of pixels by a quincunx sub-sampling pattern. Therefore, the number of pixels in one block becomes sixteen. A control signal to cause sub-sampling patterns of complementary forms every two frames is supplied from a terminal 4 to the sub-sampling circuit 3. Motion between the blocks can be detected from a difference of the pixel data of one frame by the sub-sampling.

0012Figure 2 shows sub-sampling patterns. B1, B2, B3, ... denote blocks which are located spatially at the same position and respectively belong to sequential frames F1, F2, F3, ... A number of blocks other than the blocks shown in the diagram are included in each frame. In Figure 2, ○ and Δ denote transmission pixels, and X indicates non-transmission pixels. In respective two frame periods, such as the frames F1 and F2, the sub-sampling patterns are the same. As shown by partitioning by broken lines, the pattern of two frames and the pattern of the next two frames have complementary shapes.

0013An output signal of the sub-sampling circuit 3 is supplied to a frame dropping circuit 5 to drop frames on a block unit basis, a comparing circuit 6, and a frame delay circuit 7. A control signal to control whether frames should be dropped or not, that is, to control the transmission/non-transmission of the block is supplied to the frame dropping circuit 5 through a switching circuit 8. The switching circuit 8 is controlled by a switching signal from a terminal 9 so as alternately to select input terminals a and b at one frame period.

0014A signal of "1" is always supplied to the input terminal a of the switching circuit 8. For the frame dropping circuit 5, "1" denotes that the block is transmitted. Therefore, the blocks B1, B3, B5, ... included in the first frames F1, F3, F5, ... of two sequential frames are always transmitted. In the following frames F2, F4, F6, ..., the input terminal b of the switching circuit 8 is selected and the transmission/non- transmission is adaptively controlled. That is, the transmission/non-transmission of the block are controlled on the basis of the absolute value of a frame difference between the preceding frame and the present frame.

0015The comparing circuit 6 detects a frame difference between the values of the pixels of the present frame and the preceding frame. The frame difference is converted into the absolute value by an absolute value forming circuit 10, an output signal of which is supplied to a summing circuit 11. The absolute value of the frame differences with respect to every two pixels are summed by the summing circuit 11 for one block. An output of the summing circuit 11 is supplied to a comparing circuit 12.

0016A threshold value M regarding the motion is supplied to the comparing circuit 12 from a buffering circuit 19, which will be explained below. When the summed value is equal to or smaller than the threshold value M, an output of the comparing circuit 12 is set to "0". In the other cases, the output of the comparing circuit 12 is set to "1". In other words, when the summed value is equal to or less than the threshold value M, the frame dropping process in which the block is not terminated is executed. Data corresponding to the motion between the blocks can be derived by other methods.

0017The output of the comparing circuit 12 to control the transmission/non-transmission of the block is also supplied to a detecting circuit 20 to detect a transmission ratio w for one frame period. Since the adaptive frame dropping is executed in the frames F2, F4, ... after every two frames, the transmission ratio w relative to the original data changes in each following frame. Since the total number of blocks in one frame is known, the ratio of the number of blocks which are transmitted to the total number is detected as the transmission ratio w. An output signal of the detecting circuit 20 is supplied to the buffering circuit 19. The output signal of the detecting circuit 20 is also necessary to discriminate the block which has been frame-dropped on the reception side, and is supplied as a frame dropping flag to a frame segmentation circuit 22 and is transmitted together with encoding data, which will be explained below.

0018An encoder of the ADRC is connected to the frame dropping circuit 5. The buffering circuit 19 is provided in conjunction with the encoder.

0019An output signal of the frame dropping circuit 5 is supplied to a detecting circuit 13 and a delay circuit 14. The detecting circuit 13 detects the dynamic range DR and the minimum value MIN of the block. The delay circuit 14 delays data by a time to detect the minimum value MIN and the dynamic range DR. A subtracting circuit 15 subtracts the minimum value MIN from the video data from the delay circuit 14, so that the video data from which the minimum value has been eliminated is derived from the subtracting circuit 15.

0020The output data of the subtracting circuit 15 and the dynamic range DR are supplied to a quantizing circuit 18 through delay circuits 16 and 17, respectively. A code signal DT of an n-bit length (n = 0, 1, 2, 3, or 4 bits) smaller than the original bit length (8 bits) is obtained from the quantizing circuit 18. The quantizing circuit 18 executes a quantization adapted to the dynamic range DR. That is, the video data from which the minimum value has been eliminated are divided by a quantization step Δ which is obtained by dividing the dynamic range DR into 2ⁿ equal steps. An integral value which is obtained by omitting a fraction of the quotient is used as a code signal DT. The quantizing circuit 18 can be constructed by a dividing circuit or a ROM.

0021A bit length n which is assigned to the code signal DT is determined in a manner such that a generation data amount per predetermined period, for instance, every two frames does not exceed a target value L (bits). For such buffering, the buffering circuit 19 to which the dynamic range DR is supplied is provided. A plurality of, for example, 32 sets (T1, T2, T3, T4, M) of threshold values are prepared for the buffering circuit 19 as will be explained below. The sets of the threshold values are distinguished by parameter codes Pi (i = 1, 2, 3, ..., 32). The generation data amount is set so as monotonously to decrease as the number i of the parameter code Pi increases. However, the picture quality of the reconstructed image deteriorates with decrease in generation data amount.

0022The threshold values T1 to T4 in the level direction from the buffering circuit 19 and the dynamic range DR which has been transmitted through the delay circuit 17 are supplied to a bit length deciding circuit 21. The delay circuits 16 and 17 are provided to delay the data by a time which is required to decide the threshold values by the buffering circuit 19. The dynamic range DR and the threshold values T1 to T4 (T1 &lt; T2 &lt; T3 &lt; T4) are supplied to a bit length deciding circuit 21. The bit length n to be assigned is determined on the basis of the relations of the magnitudes between the dynamic range DR and the threshold values T1 to T4.

0023The encoded outputs DR, MIN and DT of the ADRC encoder, the flag indicative of the transmission/non-transmission of the block, and the parameter codes Pi are supplied to the frame segmentation circuit 22. The transmission data is taken out to an output terminal 23. The frame segmentation circuit 22 forms the transmission data in which a sync signal has been added to the above encoded outputs. The frame segmentation circuit 22 also executes an encoding of an error correction code.

0024Although not shown, on the reproducing side, there are provided a frame desegmentation circuit, a decoder of the ADRC, a circuit to interpolate the frame dropped block by the preceding block, a circuit to interpolate the non-transmission pixels, a block desegmentation circuit, and the like. The ADRC decoder decodes the bit length n by the threshold values T1 to T4 which are designated by the parameter codes Pi and the dynamic range DR, and reconstructs the level by using the quantization step Δ according to the bit length n and the dynamic range DR and the value of the code signal DT. Further, the minimum value MIN is added to the reconstructed level.

0025Figure 3 shows an example of the buffering circuit 19. A memory (RAM) forms a frequency distribution table and an accumulative frequency distribution table. An address is supplied to the memory 31 through a multiplexer 32. The dynamic range DR is supplied as one input of the muliplexer 32 from an input terminal 33 and an address from an address generating circuit 41 is supplied as another input. An output signal of an adding circuit 34 is supplied to the memory 31. Output data of the memory 31 and an output of a multiplexer 35 are added by the adding circuit 34.

0026An output of the adding circuit 34 is supplied to a register 36. An output of the register 36 is supplied to the multiplexer 35 and a comparing circuit 37. In addition to the output of the register 36, 0 and +1 are supplied to the multiplexer 35. When an arithmetic operation of a generation data amount has been performed, a data amount li for one frame period which is generated by the ADRC encoding is obtained in an output of the register 36.

0027The comparing circuit 37 compares the generation data amount li and a target value from a switching circuit 38. An output signal of the comparing circuit 37 is supplied to a parameter code generating circuit 39 and a control signal generating circuit 48. The parameter codes Pi from the parameter code generating circuit 39 are supplied through a switching circuit 40 to the address generating circuit 41 and a register 42. The parameter codes Pi taken into the register 42 are supplied to the frame segmentation circuit 22 as mentioned above and are also supplied to a ROM 43. A table of threshold values shown in Figure 4 has been stored in the ROM 43. The ROM 43 generates sets (T1i, T2i, T3i, T4i, Mi) of the threshold values in correspondence to the parameter codes Pi which have been input as addresses. The threshold value table is constructed so as to more severely limit the generation data amount as the number of the parameter code Pi increases. That is, the values of the threshold values T1 to T4 in the level direction and the frame dropping threshold value M monotonously increase. As mentioned above, the threshold values in the level direction are supplied to the bit length deciding circuit 21 and the threshold value M regarding the motion is supplied to the comparing circuit 12.

0028The transmission ratio w which has been detected by the detecting circuit 20 is supplied to an arithmetic operating circuit 45. Assuming that a target value for the generation data amount for a two-frame period of time of F1 and F2 or the like is set to L, the arithmetic operating circuit 45 executes an arithmetic operation of L/(1 + w). The result of the arithmetic operation is supplied to a comparing circuit 46. An output of the arithmetic operating circuit 45 is used as a target value about the present frame when the mean value of the numbers of the sets of the threshold values for the preceding two-frame period is applied to the present frame.

0029A generation data amount 1′ transmitted through a register 47 is supplied to the comparing circuit 46. A comparison output of the comparing circuit 46 is supplied to the control signal generating circuit 48. The generation data amount 1′ is derived with respect to the present frame when the mean value of the numbers of the sets of the threshold values for the preceding two-frame period is applied to the present frame. The generation data amount 1′ is also supplied to a subtracting circuit 49. The subtracting circuit 49 subtracts 1′ from a target value L for the 2-frame period. (L - 1′) from the subtracting circuit 49 is supplied to an input terminal d of the switching circuit 38. A target value of L/2 is supplied to the other input terminal c of the switching circuit 38.

0030The output signal of the comparing circuit 46 is supplied to the control signal generating circuit 48.

0031The control circuit 48 generates control signals to clear the registers 36, 42 and 47, and a register 50, a signal to control the fetching of the data into the register 42, and switching signals to respectively control the switching circuits 38 and 40. The switching circuit 38 supplies the target value of L/2 to the comparing circuit 37 through the input terminal c in the initial status and upon scene change. In other cases, the switching circuit 38 supplies an output signal of the subtracting circuit 49 to the comparing circuit 37 as a target value.

0032The parameter codes Pi from the parameter code generating circuit 39 are supplied to an input terminal e of the switching circuit 40, the register 50, and an adding circuit 51. An output of the register 50 is supplied to the adding circuit 51. An output of the adding circuit 51 is supplied to a 1/2 multiplying circuit 52. The number of the mean value of the numbers Ni and Ni + 1 of the threshold values of two sequential frames is generated from the multiplying circuit 52. The mean value is supplied to an input terminal f of the switching circuit 40. The switching circuit 40 is controlled by the switching signal from the control signal generating circuit 48. When the terminal e of the switching circuit 40 is selected, the parameter codes Pi from the parameter code generating circuit 39 are supplied to the address generating circuit 41. On the other hand, when the input terminal f is selected, the parameter codes Pi indicative of the averaged threshold value number are supplied to the address generating circuit 41.

0033In the initial state or just after a scene change has occurred, buffering to control the amount of data generated for one frame period to L/2 or less is executed. That is, in the above case, as shown in Figure 3, the switching circuit 38 selects the input terminal d and the switching circuit 40 selects the input terminal 3. The operation of the buffering circuit 19 in this state will now be described with reference to a flowchart of Figure 5.

0034In the first step 61, the memory 31 and the registers 36, 42, 47, and 50 are cleared to zero. To clear the memory 31 to zero, the multiplexer 32 selects the address generated from the address generating circuit 41 and the output of the adding circuit 34 is always set to zero. The address changes to (0, 1, 2, ..., 255) and the zero data is written into all of the addresses in the memory 31.

0035In the next step 62, a frequency distribution table of the dynamic range DR of a certain frame as a unit period of time when the buffering is executed is formed in the memory 31. The multiplexer 32 selects the dynamic range DR from the terminal 33. The multiplexer 35 selects +1. Therefore, at the end of one frame period, the generation frequency of each DR is stored into each address in the memory 31 corresponding to the dynamic range DR. In the frequency distribution table in the memory 31, an axis of abscissa indicates the DR and an axis of ordinate represents the frequency as shown in Figure 6A.

0036The frequency distribution table is converted into the accumulative frequency distribution table (step 63). When the accumulative frequency distribution table is formed, the multiplexer 32 selects the address from the address generating circuit 41 and the multiplexer 35 selects the output of the register 36. The address is sequentially decreased from 255 to 0. The read-out output of the memory 31 is supplied to the adding circuit 34 and is added to the content in the register 36 by the adding circuit 34. The output of the adding circuit 34 is written into the same address as the read address of the memory 31. The content in the register 36 is updated to the output of the adding circuit 34. In the initial status in which the address in the memory 31 is set to 255, the register 36 has been cleared to zero. When the frequencies have been accumulated with respect to all of the addresses in the memory 31, an accumulative frequency distribution table shown in Figure 6B is formed in the memory 31.

0037A generation data amount li when the sets (T1i, T2i, T3i, T4i) of the threshold values in the level direction have been applied to the accumulative frequency distribution table is calculated (step 64). When the generation data amount li is calculated, the multiplexer 32 selects the output of the address generating circuit 41 and the multiplexer 35 selects the output of the register 36. The parameter code generating circuit 39 generates a parameter code which sequentially changes from P1 to P32. The parameter codes Pi are supplied to the address generating circuit 41 through the switching circuit 40 and the addresses corresponding to the threshold values of (T1i, T2i, T3i, T4i) are sequentially generated.

0038Values A1, A2, A3 and A4 which have been read out of the addresses corresponding to the threshold values are accumulated by the adding circuit 34 and the register 36. The accumulative value (A1 + A2 + A3 + A4) corresponds to the generation data amount li when the sets of the threshold values designated by the parameter codes pi have been applied. That is, in the accumulative frequency distribution table shown in Figure 6B, a value in which the number (16) of pixels in the block has been multiplied to the total value (A1 + A2 + A3 + A4) of the values A1, A2, A3, and A4 which had been read out of the addresses respectively corresponding to the threshold values T1, T2, T3, and T4 is equal to the generation data amount (bit length). Since the number of pixels is omitted in the buffering circuit 19 shown in Figure 3.

0039The generation data amount li is compared with the target value L/2 (step 65). The output of the comparing circuit 37 which is generated when (Ai ≦ L/2) is satisfied is supplied to the parameter code generating circuit 39. The increment of the parameter code Pi is stopped. The parameter code Pi from the register 42 and the set of the threshold values generated from the ROM 43 are generated (step 66).

0040If (li ≦ L/2) is not satisfied in discriminating step 65 in the comparing circuit 37, the parameter codes Pi are changed to the next parameter codes Pi + 1 and the address corresponding to Pi + 1 is generated from the address generating circuit 41. In a manner similar to the above, a generation data amount li + 1 is calculated and compared with the target value L/2 by the comparing circuit 37. The above operations are repeated until (li ≦ L/2) is satisfied.

0041The operation of the embodiment will be described with reference to Figure 7. In the initial state in which the image data of the first frame F1 shown in Figure 2 is input, the buffering to reduce the amount li of data which is generated for the frame F1 by the above operation to L/2 or less is executed. The encoding of the two-dimensional ADRC is executed by the threshold values T1 to T4 in the level direction at that time (step 71). N1 denotes the number of the threshold value used in the frame F1, that is, the number which coincides with the parameter code Pi of the threshold value used.

0042In the next frame F2, the frame dropping process is executed on a block unit basis by the threshold value M regarding the motion of N1 (step 72). The transmission ratio w of the frame F2 after completion of the frame dropping process is detected by the detecting circuit 20 (step 73). The data of the frame F2 after completion of the frame dropping process is ADRC encoded (step 74). The threshold value at that time is set so as to reduce the generation data amount 12 to L/2 or less in a manner similar to the frame F1.

0043With respect to the frames F1 and F2, the target value L for the 2-frame period is set to the half value, thereby setting the target value L/2 for the one frame period. For the frame F2, however, an amount of data to be transmitted is reduced by sub-sampling. Thus, as shown in Figure 8, the number N1 of the threshold value about the frame F1 is large and the number N2 of the threshold value regarding the frame F2 is fairly small.

0044After an accumulative frequency distribution table was formed with regard to the next frame F3, a generation data amount 13′ when the encoding has been executed by the mean value N3 of the numbers N1 and N2 of the threshold values of the two preceding frames F1 and F2 is obtained (step 75). The generation data amount 13′ is supplied to the comparing circuit 46 through the register 47. The comparing circuit 46 executes a comparing operation of (13′ ≦ L/(1 + w2)) (step 76).

0045When the above relation is satisfied, the data of the frame F3 is encoded by the threshold value of the number N3 (step 77). On the contrary, when the above relation is not satisfied, it is determined that a scene change occurs. The buffering so as to reduce the generation data amount 13 regarding the frame F3 into L/2 or less is performed and the data of the frame F3 is encoded by the threshold value determined by the buffering (step 81). The above process is similar to those in the initial status in step 71. The operations similar to those in step 71 and the subsequent steps are repeated hereinbelow.

0046When the relation in step 76 is satisfied, the frame dropping process regarding the frame F4 is executed by the threshold value M3 included in the number N3 of the threshold value (step 78). A ratio w4 of the transmission data of the frame F4 after the frame dropping is detected (step 79).

0047In the buffering about the frame F4, (L - 13′) formed by the subtracting circuit 49, is used as a target value. That is, in the preceding frame F3, since the data of 13′ has been generated for the target value L for the 2-frame period, the generation data amount of the frame F4 is suppressed to a value which is equal to or less than the remaining data amount of (L - 13′). The data of the frame F4 is encoded by the number N4 of the threshold value determined by the buffering (step 80). The next process of step 80 is similar to that in step 75 and processes similar to those mentioned above are repeated.

0048In the embodiment in which the above operations are executed, when a scene change does not occur, as shown in Figure 8, the number of the threshold value can be converged to a mean value shown by a broken line for the frame F3 and subsequent frames.

0049In the above description, the dynamic range DR and the minimum value MIN are transmitted in order to transmit the dynamic range information. However, the maximum value MAX or the quantization step width can be also transmitted as an additional code in place of the dynamic range DR.

0050On the other hand, in the above embodiment, the encoding process has been applied to the pixel data themselves of the block after each 2-frame period. However, the encoding can be also performed to a difference (residual) between each pixel data of the preceding frame and each pixel data of the post frame. That is, in such a case, the data corresponding to the pixels in each block of the post frame has differential value.

0051The invention, further, can use a block encoding of a discrete cosine transform (DCT) or the like, instead of ADRC. In the case of DCT, encoding using the above differential value in place of the pixel data is effective.

0052Since the compression encoding of the ADRC or the like is executed for a two-dimensional block, the necessary memory capacity can be reduced as compared with a three-dimensional block. On the other hand, since there is used a hybrid construction of the frame dropping on a block unit basis and the compression encoding, the compression ratio can be raised.