JP3618783B2 - Image encoding method and image encoding apparatus - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to an image coding method and an image coding apparatus for coding a sequence of image data in combination with intra-coded and inter-picture encoding.
[0002]
[Prior art]
Since the amount of data of image information becomes enormous, various compression encoding methods have been proposed and studied. For example, a combination of an intra-frame coding scheme and an inter-frame coding scheme is considered promising.
[0003]
The intra-frame coding method compresses information by utilizing the characteristic of an image that neighboring pixels in the frame have similar brightness and color. In an actual image, since many portions such as the sky and the wall have the same brightness and color, information can be compressed to about 1/5 to 1/10.
[0004]
Interframe coding uses the fact that temporally adjacent frames become similar images in moving images, and only encodes information on differences between frames and transmits them. More specifically, in a normal moving image, the patterns of adjacent frames are basically similar, although there are some movements and deformations. Using this point, first, the similarity (movement, color, brightness, etc.) between the frame to be compression-encoded and a frame close to it (for example, the immediately preceding frame) is calculated, and the calculation is performed. Based on the result, a predicted value of a frame to be encoded is calculated from the adjacent frames. Then, the difference between the frame to be encoded and the predicted value is encoded and transmitted.
[0005]
For example, in a moving image in which only a person is shown and the person is moving to the right, only the difference information of the moving person needs to be encoded, and a high compression rate can be realized. If a motion compensation prediction method is added to this, the movement information of the movement will increase, but the person image will almost match before and after the movement, so the difference value of each pixel of the person image will also be very small, and overall higher compression Rate can be achieved.
[0006]
Note that intra-frame coding and inter-frame coding may be selected not for the entire frame but for some blocks of the frame. That is, part of the frame may be intra-coded and the rest may be inter-coded.
[0007]
Since the inter-frame coding system has a property that a transmission error propagates, it is necessary to insert the intra-frame coding system at an appropriate interval. This is called refresh.
[0008]
[Problems to be solved by the invention]
The intra-frame coding method generally generates more data than the inter-frame coding method, so the amount of data increases rapidly due to refresh, and the image quality may be greatly degraded due to the balance with the transmission rate and the like. there were.
[0009]
It is an object of the present invention to provide an image encoding method and an image encoding apparatus that do not cause such inconvenience.
[0010]
[Means for Solving the Problems]
An image encoding apparatus according to another aspect of the present invention for solving the above-described problem includes an input means for inputting image data to be encoded, image data obtained by decoding the encoded image data of the previous screen, and the encoding A generation unit that obtains a motion vector from the target image data, generates predicted image data for the encoding target image data according to the motion vector, and a first area in one screen of the encoding target image data Intra-screen coding is forcibly applied, and intra-frame coding and inter-frame coding using predicted image data generated by the generating means are applied to the second area other than the first area in the one screen. encoding means for block coding using a manner, in the image coding apparatus and a control means for moving in a predetermined screen every while overlapping the first area, the control hand Further wherein the predictive picture data for the coding object image data of the overlap area of the first area is, depending whether the previous ones obtained from images included in the first area of the screen to the motion vector When the predicted image data is obtained from an image included in the first area of the previous screen , the second area is not subjected to the forcible intra-frame encoding on the encoding target image data. Control is performed so as to execute the same encoding process.
The image coding method of one invention for solving the above-mentioned problems, an input step of inputting the coded image data, and image data obtained by decoding the image data of the previous screen encoded the A generation step of obtaining a motion vector from the encoding target image data, generating predicted image data for the encoding target image data according to the motion vector, and a first area in one screen of the encoding target image data Forcibly intra-screen encoding, intra-screen encoding for the second area other than the first area in the one screen, and inter-screen encoding using the predicted image data generated in the generation step, the a adaptively a coding step of block coding using the image coding method and a control step of moving a predetermined amount in a predetermined screen every while overlapping the first area, before Control step, further wherein the predictive picture data for the coding object image data of the overlap area in the first area, wherein whether said before those obtained from images included in the first area of the screen motion vector If the predicted image data is obtained from an image included in the first area of the previous screen, the compulsory intra-screen encoding is not performed on the encoding target image data, and the first Control is performed so that encoding processing similar to that in the two areas is executed.
[0012]
【Example】
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
[0013]
FIG. 1 shows a schematic block diagram of an embodiment of the present invention. In FIG. 1, 10 is an input terminal for inputting an analog image signal to be encoded, 12 is an A / D converter for digitizing the analog image signal from the input terminal 10, and 14 is an A / D converter 12. It is a block division circuit that divides output image data into blocks and outputs them in the order of blocks.
[0014]
16 is a subtracter for subtracting the prediction value in the prediction differential encoding from the output of the block dividing circuit 14, 18 is a switch for selecting the output of the block dividing circuit 14 (a contact) or the output of the subtractor 16 (b contact), Reference numeral 20 denotes an orthogonal transformation circuit that performs orthogonal transformation (for example, discrete cosine transformation) on an output of the switch 18 in units of a predetermined size, and 22 denotes a quantization circuit that quantizes the transformation coefficient output from the orthogonal transformation circuit 20. is there.
[0015]
Reference numeral 24 denotes an inverse quantization circuit that inversely quantizes the output of the quantization circuit 22, reference numeral 26 denotes an inverse orthogonal transform circuit that performs inverse orthogonal transform on the output of the inverse quantization circuit 24, and reference numeral 28 denotes an output of the inverse orthogonal transform circuit 26 to the frame An adder that adds a prediction value during inter-coding, and “0” during intra-frame coding, 30 is an image memory that temporarily stores image data decoded by the circuits 24, 26, and 28, and 32 is a block The motion vector detection circuit 34 detects a motion vector from the image data of the current frame from the dividing circuit 14 and the image data of the previous frame from the image memory 30, according to the motion vector detected by the motion vector detection circuit 32, This is a motion compensation circuit that compensates motion of image data of the previous frame from the image memory 30. The output of the motion compensation circuit 34 becomes a predicted value of interframe coding, and is applied to the adder 28 via the subtracter 16 and the switch 36.
[0016]
The switch 36 is switched in conjunction with the switch 18, and is connected to the contact a to apply the value “0” to the adder 28 for intra-frame coding, and connected to the contact b for inter-frame coding. The output of the motion compensation circuit 34 is applied to the adder 28. The switch 36 is switched in conjunction with the switch 18.
[0017]
Reference numeral 38 denotes a prediction region detection circuit which detects a region where refresh should be canceled from the detection output of the motion vector detection circuit 32 and outputs a refresh cancellation signal. The detailed operation of this circuit 38 will be described later.
[0018]
40 is a determination circuit that compares the output of the block dividing circuit 14 and the output of the subtractor 16 to determine which of the inter-frame coding and the intra-frame code has the higher coding efficiency, and 42 indicates the set values W and S. And a refresh control circuit that controls refresh according to the refresh release signal from the prediction region detection circuit 38. Both the output of the determination circuit 40 and the output of the refresh control circuit 42 are applied to the OR circuit 44, and the output of the OR circuit 44 switches and controls the switches 18 and 36.
[0019]
46 is a variable length coding circuit that performs variable length coding on the output of the quantization circuit 22, 48 is a buffer memory that buffers the output of the variable length coding circuit 46 in accordance with the transmission rate of the transmission path, and 50 is A transmission interface circuit 52 converts the output of the buffer memory 48 into a predetermined transmission format, and 52 is an output terminal for connecting the output of the transmission interface circuit 50 to the transmission path.
[0020]
First, the basic flow of image signal processing will be described. The A / D converter 12 digitizes the analog image signal input from the input terminal 10. The block dividing circuit 14 divides the output of the A / D converter 12 into blocks of a (for example, 8) pixels in the horizontal direction and b (for example, 8) lines in the vertical direction. The output of the block dividing circuit 14 is applied to the contact a of the switch 18, the subtracter 16, the motion vector detection circuit 32, and the determination circuit 40.
[0021]
The subtracter 16 subtracts the predicted value (output of the motion compensation circuit 34) from the output of the block dividing circuit 14, and outputs the prediction error data to the b contact of the switch 18. The switch 18 is connected to the a contact or the b contact in block units according to the output of the OR circuit 44. When switch 18 is connected to contact a, switch 36 is also connected to contact a, and when switch 18 is connected to contact b, switch 36 is also connected to contact b. When the switches 18 and 36 are connected to the a contact, intra-frame encoding is performed, and when the switches 18 and 36 are connected to the b contact, inter-frame encoding is performed.
[0022]
The orthogonal transform circuit 20 performs orthogonal transform (for example, discrete cosine transform) on the image data (original image data or prediction error data) selected by the switch 18 for each block, and the quantization circuit 22 quantizes the transform coefficient data. To do.
[0023]
The inverse quantization circuit 24 inversely quantizes the output of the quantization circuit 22, and the inverse orthogonal transform circuit 26 inversely orthogonally transforms the output of the inverse quantization circuit 24. The switch 36 is connected to the a contact in the intra-frame coding block, and is connected to the b contact in the inter-frame coding block. Thus, the adder 28 outputs the output of the inverse orthogonal transform circuit 26 as it is in the intra-frame coding block, and adds the predicted value to the output of the inverse orthogonal transform circuit 26 in the inter-frame coding block and outputs it. To do.
[0024]
The image memory 30 delays the output of the adder 28 by one frame period and outputs it to the motion vector detection circuit 32 and the motion compensation circuit 34. The output of the block dividing circuit 14 is also supplied to the motion vector detection circuit 32. The motion vector detection circuit 32 detects a motion vector in units of encoded blocks from both signals and predicts the detection result with the motion compensation circuit 34. The data is output to the area detection circuit 38. The motion compensation circuit 34 moves the image data of the previous frame from the image memory 30 so as to cancel the motion of the image in accordance with the motion vector from the motion vector detection circuit 32, that is, compensates for motion, and uses it as a predicted value of the current frame. This is supplied to the b contact of the subtractor 16 and the switch 36.
[0025]
The variable length coding circuit 46 performs variable length coding on the output of the quantization circuit 22, and the output is temporarily stored in the buffer memory 48. The transmission interface circuit 50 reads data from the buffer memory 48 in synchronization with the transmission clock, and outputs the data to the output terminal 52 in a predetermined transmission format. Although not shown, this transmission format includes motion vector information detected by the motion vector detection circuit 32, a transmission synchronization signal, an error correction code, and the like.
[0026]
Next, operations of the prediction region detection circuit 38, the determination circuit 40, and the refresh control circuit 42 will be described in detail.
[0027]
The refresh control circuit 42 controls the “0” (ie, L) signal or “1” (“1”) in accordance with the control values W and S that define the refresh width and the shift amount, respectively, and the refresh release signal from the prediction region detection circuit 38. That is, H) signal is output. In this embodiment, refresh is performed by partially performing intra-frame coding over several frames rather than completing refresh in one frame. Even if the block or pixel is to be refreshed with the control values W and S, it is not necessary to perform intra-frame coding for a portion that does not need to be refreshed, such as intra-frame coding in the previous frame that is a predicted image. . The prediction region detection circuit 38 detects such a block that does not need to be refreshed and outputs a refresh release signal. Even if the refresh control circuit 42 is within the range of the refresh width control value W, the refresh control circuit 42 stops the refresh of that portion in response to the refresh release signal from the prediction region detection circuit 38. As a result, an increase in the amount of generated data due to intra-frame coding can be minimized while suppressing transmission error transmission.
[0028]
Even if the prediction region detection circuit 38 is not provided, if the control values W and S are set appropriately, a sudden increase in the amount of generated data at the time of refresh can be suppressed. However, by providing the prediction region detection circuit 38, An increase in the amount of generated data due to intra-frame coding for refresh can be minimized.
[0029]
This will be specifically described. In the normal encoding process, the refresh control circuit 42 outputs a “0” signal or an L signal. Accordingly, the output of the OR circuit 44 coincides with the output of the determination circuit 40, and the switches 18 and 36 are controlled to be switched by the determination circuit 40.
[0030]
For the block to be refreshed, the refresh control circuit 42 outputs a “1” signal or an H signal. As a result, the output of the OR circuit 44 becomes “1”, and the switches 18 and 36 are both connected to the “a” contact and forced to perform intra-frame coding.
[0031]
With reference to FIG. 2, the interrelationship between areas to be refreshed between frames will be described. In FIG. 2, in the first frame, an area having a set width of W is refreshed (that is, intra-frame encoded), and in the next second frame, a control value W shifted by the control value S in the horizontal direction. The width area is refreshed. The same shift operation is repeated for the third and subsequent frames, and the refresh of one screen is completed in N frames.
[0032]
The relationship between W and S will be described with reference to FIG. When the range of the width W at the left end is refreshed in the first frame, the remaining area is encoded by intra-frame encoding or inter-frame encoding, and thus generally may include an error propagating to the next frame. There is sex. If inter-frame motion compensation is not used, it is sufficient to refresh the range of width W shifted to the right by W in the second frame. However, when inter-frame motion compensation is used, an error is propagated by a distance corresponding to the motion compensation capability. Therefore, it is necessary to refresh this range in which the error can propagate again in the next frame. Specifically, W and S may be set as in the following equation.
[0033]
W ≧ M + 1
1 ≦ S ≦ W−M
Here, M is an error propagation possible distance (horizontal width). Each value of M, W, and S is a block unit or a pixel unit.
[0034]
As shown in FIG. 3, in the second frame, the refresh area moves to the right by S with respect to the first frame. However, if W and S satisfy the above conditional expression, the refresh frame is refreshed in the first frame and the second frame. Since the areas overlap by M or more, the portion that may propagate the error from the portion that is not refreshed in the first frame is refreshed again in the second frame, and the error does not propagate to the area refreshed in the first frame. .
[0035]
However, when attention is paid to the error propagation possible range (overlap area) of the second frame, the motion compensated prediction signal in this error propagation possible range is refreshed in the first frame (or intra-frame coded). If it is based on (image), there is no error propagation.
[0036]
For convenience of explanation, it is assumed that the error propagation possible range is 2 blocks in the horizontal direction (1 block is 8 pixels × 8 lines) as shown on the right side of FIG. (N) indicates a block encoded in the normal mode in the second frame, and (R) indicates a block encoded in the refresh mode in the second frame. (M, n) is a motion vector for motion compensation prediction, m represents the number of pixels in the horizontal direction, and n represents the number of lines in the vertical direction. Assuming that the frame to be encoded and its prediction frame are overlapped at the same position, the block to be encoded is encoded with the image on the left side by m pixels when m is negative, and m when m is positive. Only the pixels are blocked by the image signal on the right side. As for n, the block to be encoded is encoded with the upper image by n lines as a predicted value when n is negative, and is blocked by the lower image signal by n lines when n is positive. The
[0037]
In the example shown in FIG. 3, the upper right block 60 is predicted for motion compensation with a motion vector of −4 pixels horizontally and 0 lines vertically, and the lower block 62 is 21 pixels horizontally and vertically. Motion compensation prediction is performed with a motion vector of 0 lines. Therefore, since the block 60 of the second frame is predicted from the image on the left side of the position of the block 60 in the first frame, that is, the image included in the refresh area of the first frame, error propagation from the first frame is performed. There is nothing to do. That is, the block 60 of the second frame does not need to be refreshed.
[0038]
The prediction area detection circuit 38 detects from the output of the motion vector detection circuit 32 whether or not the predicted area is refreshed in the previous frame. Apply to. When this refresh cancellation signal is input, the refresh control circuit 42 outputs a “0” signal during the processing period of this block regardless of the control values W and S. As a result, the switches 18 and 36 are switched and controlled exclusively by 40 outputs, and operate in the normal mode.
[0039]
In block 62, the horizontal motion vector is +21 and is predicted from the unrefreshed image portion of the first frame, so an error may propagate. The prediction region detection circuit 38 detects the possibility of error propagation and does not output a refresh release signal. As a result, the refresh control circuit 42 outputs a “1” signal according to the control values W and S. As a result, the block 62 is refreshed according to the control values W and S.
[0040]
The prediction region detection circuit 38 uses the prediction motion vector from the motion vector detection circuit 32 in the overlap area of the frame to be encoded, and cancels the refresh for the block predicted from the area refreshed by the prediction frame. A signal is output, and a block that is predicted from an area that is not refreshed in the predicted frame operates so as not to output a refresh release signal. In regions other than the overlap area, the prediction region detection circuit 38 does not output a refresh release signal.
[0041]
For the remaining blocks in the error propagation possible range, the refresh control circuit 42 performs the same processing according to the control values W and S and the refresh release signal, and refreshes the minimum necessary blocks. The subsequent frames are refreshed in the same manner.
[0042]
With these operations, in this embodiment, the minimum necessary range is refreshed, and an increase in the amount of data due to the refresh can be suppressed and image quality deterioration can be reduced, so that overall image quality can be improved.
[0043]
In the above description, the refresh area in the form of a stripe extending vertically is moved horizontally. However, as shown in FIG. 4, the refresh area may be moved vertically using a stripe extending horizontally. . In this case, the prediction region detection circuit 38 generates a refresh release signal exclusively according to the motion vector in the vertical direction.
[0044]
In the present embodiment, the inter-frame motion compensation coding has been described, but it is obvious that the same applies to the inter-field motion compensation. The same applies to inter-field / inter-frame motion compensation for each motion compensation range.
[0045]
Further, although predictive coding using the previous frame as a predicted frame has been described as an example, the present invention clearly shows that the predicted frame may be two frames before or three frames before. That is, the prediction frame is not limited to one frame before, and is not limited to being the previous frame. For example, the present invention can be applied to predictive coding in which a subsequent frame, or the previous and subsequent frames are predicted frames, and further predictive coding in which these are combined.
[0046]
【The invention's effect】
As can be easily understood from the above description, according to the present invention, an increase in the amount of data due to refresh can be reduced, and further error propagation can be prevented, so that encoding efficiency and image quality can be improved. Can do.
[Brief description of the drawings]
FIG. 1 is a configuration block diagram of an embodiment of the present invention.
FIG. 2 is an explanatory diagram of a refresh operation according to the present embodiment.
FIG. 3 is an explanatory diagram of the effect of a refresh area shift and a refresh release signal.
FIG. 4 is another example of setting a refresh area.
[Explanation of symbols]
10: input terminal 12: A / D converter 14: block dividing circuit 16: subtractor 18: switch 20: orthogonal transformation circuit 22: quantization circuit 24: inverse quantization circuit 26: inverse orthogonal transformation circuit 28: adder 30 : Image memory 32: Motion vector detection circuit 34: Motion compensation circuit 36: Switch 38: Predictive region detection circuit 40: Determination circuit 42: Refresh control circuit 44: OR circuit 46: Variable length encoding circuit 48: Buffer memory 50: Transmission interface circuit 52: output terminal

Claims (4)

Input means for inputting image data to be encoded;
A motion vector is obtained from the image data obtained by decoding the encoded image data of the previous screen and the encoding target image data, and predicted image data for the encoding target image data is generated according to the motion vector Generating means to
The first area in one screen of the encoding target image data is forcibly encoded in the screen, the second area other than the first area in the one screen is encoded in the screen, and the generating means Encoding means for adaptively using inter-frame encoding using predicted image data generated by
An image encoding device having control means for moving each predetermined screen while overlapping the first area,
The control means further determines whether the predicted image data for the image data to be encoded in the overlap area in the first area is obtained from an image included in the first area of the previous screen. Judging according to the vector, when the predicted image data is obtained from the image included in the first area of the previous screen, without performing the forced intra-screen coding on the encoding target image data, An image encoding apparatus that controls to execute an encoding process similar to that in the second area. The width W with respect to the moving direction of the first area is:
W ≧ M + 1
M: a value satisfying a maximum search range amount capable of motion compensation in the moving direction,
The control means includes
1 ≦ S ≦ (WM)
S: Data indicating a movement amount S set to a value satisfying the movement amount of the first area to be moved for each predetermined screen is input, and the first area is moved based on the input data. The image encoding device according to claim 1. An input process for inputting the image data to be encoded;
A motion vector is obtained from the image data obtained by decoding the encoded image data of the previous screen and the encoding target image data, and predicted image data for the encoding target image data is generated according to the motion vector Generating process to
The first area in one screen of the encoding target image data is forcibly encoded in the screen, and the second area other than the first area in the one screen is encoded in the screen and the generation step An encoding step of adaptively using inter-frame encoding using predicted image data generated in step B, and
An image coding method and a control step of moving a predetermined amount the first area for each overlapped allowed while predetermined screen,
The control step further determines whether the predicted image data for the image data to be encoded in the overlap area in the first area is obtained from an image included in the first area of the previous screen. Judging according to the vector, when the predicted image data is obtained from the image included in the first area of the previous screen, without performing the forced intra-screen coding on the encoding target image data, An image encoding method, wherein control is performed so as to execute an encoding process similar to that in the second area. The width W with respect to the moving direction of the first area is W ≧ M + 1.
M: a value satisfying a maximum search range amount capable of motion compensation in the moving direction,
In the control step,
1 ≦ S ≦ (WM)
S: Data indicating a movement amount S set to a value satisfying the movement amount of the first area to be moved for each predetermined screen is input, and the first area is moved based on the input data. The image encoding method according to claim 3.

Images