KR100203697B1 - Improved image coding system having functions for controlling generated amount of coded bit stream - Google Patents

Abstract

The present invention calculates the complexity of an image to be currently encoded based on quantization parameter information determined in units of macroblocks based on the data fullness state information of the output transmission buffer, and based on the calculation result, The present invention relates to a video encoding system having a bit generation amount adjustment function for adaptively adjusting a bit generation amount after encoding by adjusting a weight for a component. To this end, the present invention relates to each macroblock of a previous frame that has been encoded. Average value calculating means for calculating an average quantization parameter value of a previous frame by averaging each quantization parameter value generated for each of the previous frames; The average quantization parameter value calculated in the previous frame is referred to as the complexity of the frame to be currently encoded by the encoding means, and adaptively limits the bandwidth during low pass filtering of each DCT transform coefficient block generated through the discrete cosine transform. The apparatus further includes weight generation means for determining a weight corresponding to the average quantization parameter value of the previous frame calculated from among the plurality of preset weights and providing the weighted value to the encoding means, wherein the encoding means comprises: before quantization for each DCT transform coefficient block. By restricting the pass frequency band of each DCT transform coefficient block based on the determined weight, and selectively filtering the high frequency components of each DCT transform coefficient block, the amount of bits generated after encoding without increasing the excessive step size in the quantized step Can effectively control .

Description

Image Coding System with Bit Rate Control

1 is a block diagram of a video encoding system having a bit generation amount adjusting function according to an embodiment of the present invention.

2 is a diagram showing a weight determination region determined based on the complexity of an 8x8 pixel block as an example according to the present invention.

* Explanation of symbols for main parts of the drawings

100170: frame memory 110: subtractor

120: image coding block 130: entropy coding block

140: transmission buffer 150: video decoding block

160: adder 180: current frame prediction block

210: QP average value calculation block 220: weighting block

The present invention relates to a video encoding system for compressing and encoding a video signal. More particularly, when encoding a video in a video encoder that compresses and encodes a video signal, each macroblock unit is based on data fullness state information of an output transmission buffer. The present invention relates to a video encoding system having a bit generation amount adjustment function suitable for adaptively adjusting the amount of generated bits after encoding with reference to a variation of an input video signal predicted based on quantization parameter information determined as.

As is well known in the art, the transmission of discrete video signals can maintain better image quality than analog signals. When a video signal consisting of a series of image frames is represented in digital form, a significant amount of data must be transmitted, especially for high quality televisions (aka HDTVs). However, since the usable frequency range of the conventional transmission channel is limited, in order to transmit a large amount of digital data, it is necessary to compress the transmitted data and reduce the amount thereof. Among the various compression methods for compressing data, hybrid coding which combines stochastic coding with temporal and spatial compression is known to be the most efficient, and these methods are already established by the World Standards Organization. It is widely disclosed in recommendations such as MPEG-1 and MPEG-2.

Most hybrid coding techniques use motion compensated DCPM (differential pulse code modulation), two-dimensional DCT (discrete cosine transform), quantization of DCT coefficients, variable-length coding (VLC), and the like. The motion compensation DPCM determines a motion of an object between a current frame and a previous frame, and predicts a current frame according to the motion of the object to generate a difference signal representing a difference between the current frame and a predicted value. This can be done for example by Staffan Ericsson's Fixed and Adaptive Predictors for Hybrid Predictive / Transform Coding, IEEE Transactions on Communication, COM-33, NO.12 (December 1985), or A motion Compensated Interframe Coding Scheme by Ninomiy and Ohtsuka. for Television Pictures, IEEE Transactions on communication, COM-30, NO.1 (January 1982).

In general, two-dimensional DCT converts digital image data blocks, for example, 8 x 8 blocks, into DCT conversion coefficients by using or removing spatial redundancy between image data. This technique is described in Chen and Pratt's Scene Adaptive Coder, IEEE Transactions on communication, COM-32, NO.3 (March, 1984). These DCT transform coefficients are derived from quantizers, zigzag scans, and VLC. By processing, the amount of data to be transmitted can be effectively reduced (or compressed).

More specifically, the motion compensation DPCM predicts the current frame from the previous frame according to the motion of the object estimated between the current frame and the previous frame. The estimated motion may be represented by a two-dimensional motion vector representing the displacement between the previous frame and the current frame.

Typically, there are several approaches to estimating the displacement of an object. These are generally classified into two types, one of which is a block-based motion estimation method using a block matching algorithm, and the other is a minimum-unit motion estimation method using a pixel circulation algorithm.

In the motion estimation method for estimating the displacement of an object as described above, the displacement is obtained for each pixel by using the pixel-based motion estimation method. This method has the assumption that the pixel value can be estimated more accurately and the scale change (e.g., zooming, a movement perpendicular to the image plane) can be easily handled, while the motion vector is determined for each pixel. Since a large amount of motion vectors occurs, it is impossible to transmit substantially all of the motion vectors to the receiver.

In addition, in block-by-block motion estimation, a block having a predetermined size of a current frame is moved by one pixel in a search range of a previous frame and compared with corresponding blocks to determine an optimal matching block having a minimum error value. From this, the interframe displacement vector (the degree of block movement between frames) for the entire block is estimated for the current frame to be transmitted. Here, in determining the similarity between two corresponding blocks between the current frame and the previous frame, the average absolute difference, the mean square difference, etc. are mainly used, as is well known in the art.

On the other hand, the image bit stream encoded by the encoding technique as described above, that is, encoding techniques such as motion compensation DCPM, two-dimensional DCT, DCT coefficient quantization, and VLC (or entropy encoding) is transmitted to the output side of the image encoding system. The next transmission point stored in the buffer is sent to the transmitter for transmission to the remote destination. At this time, the transfer time here is related to the size (i.e. capacity) and transfer rate of the transfer buffer and is controlled so that no malfunction (data overflow or data underflow) in the transfer buffer occurs. .

More specifically, the amount of bits generated for each frame during encoding is changed due to various factors (for example, the complexity of the image). In view of this, in the image encoding system, the average bit rate may be kept constant. Control of the output buffer. That is, the video encoding system checks the amount of bits generated before the frame currently encoded based on the data full state information of the output side transmission buffer and adjusts the amount of bits to be allocated in the current frame. In other words, in the conventional typical video encoding system, the amount of generated bits in the encoding system is adjusted by adjusting the quantization step size (QP) based on the data fullness state information of the side-side transmission buffer, that is, the amount of generated bits up to the previous stage. As much as possible, the bit generation amount is controlled by increasing the quantization step size to reduce the bit generation amount, and in the opposite case, the bit generation amount is controlled by decreasing the quantization step size to increase the bit generation amount.

However, as described above, the conventional method of adjusting the bit generation amount by adjusting the quantization step size based on the data fullness state information of the output side transmission buffer is performed when encoding and transmitting video data corresponding to each frame at the same data rate. In the case where the image to be encoded is complex (a large amount of high frequency components are generated), a large amount of bits is generated, thereby increasing the quantization step size, resulting in a serious deterioration of image quality in the reproduced image. The high frequency component generated here is a component that is substantially insensitive to human vision (a component that hardly affects the image quality of a reproduced video).

Accordingly, the present invention is to solve the above-described problems of the prior art, and based on the data fullness state information of the output transmission buffer to determine the complexity of the image to be currently encoded based on the quantization parameter information determined in each macro block unit. It is an object of the present invention to provide a video encoding system having a bit generation amount adjustment function capable of adaptively adjusting a bit generation amount after encoding by calculating and adjusting a weight for a high frequency component during quantization based on the calculation result.

In order to achieve the above object, the present invention includes a discrete cosine transform, quantization and entropy encoding on the difference signal between the input current frame and the prediction frame obtained through motion estimation and compensation using the current frame and the reconstructed previous frame. And a bit stream encoded by compression encoding through an encoding means, wherein the quantization is a bit generation amount adjusting function of adjusting a step size based on a quantization parameter based on the full state information of the bit stream stored in an output buffer. An image encoding system having: an average value calculating means for calculating an average quantization parameter value of the previous frame by averaging each quantization parameter value generated for each of the macroblocks of the immediately preceding frame to be encoded; The average quantization parameter value calculated in the previous frame is referred to as the complexity of the frame to be currently encoded by the encoding means, and the bandwidth is adaptive when low pass filtering of each DCT transform coefficient block generated through the discrete cosine transform. The apparatus may further include weight generation means for determining a weight corresponding to the calculated average quantization parameter value of the previous frame among the plurality of preset weights to be provided to the encoding means, wherein the encoding means comprises: each DCT transform; The bit generation amount adjusting function of filtering the high frequency component of each DCT transform coefficient block by limiting the pass frequency band of each DCT transform coefficient block based on the determined weight before quantization of coefficient blocks. Providing an image encoding system All.

The above and other objects and various advantages of the present invention will become more apparent from the preferred embodiments of the present invention described below with reference to the accompanying drawings by those skilled in the art.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram of a video encoding system having a bit generation amount adjusting function according to a preferred embodiment of the present invention. As shown in the figure, the image encoding system of the present invention includes a first frame memory 100, a subtractor 110, an image encoding block 120, an entropy encoding block 130, a transmission buffer 140, and image decoding. A block 150, an adder 160, a second frame memory 170, a current frame prediction block 180, a quantization parameter average value calculation block 210, and a weight generation block 220 are included.

Referring to FIG. 1, the input current frame signal is provided to the subtractor 110 and the current frame prediction block 180 through the next line L11 stored in the first frame memory 100, respectively.

First, the subtractor 110 performs motion compensated predicted motion on the moving object provided from the current frame prediction block 180 described later through the line L19 from the current frame signal provided from the first frame memory 100 through the line L11. Subtract the current frame signal and, as a result, the data, i.e., the error signal representing the differential pixel value, is obtained using a discrete cosine transform (DCT) and one of the quantization methods well known in the art through the image coding block 120. By encoding a series of quantized DCT transform coefficients. At this time, the quantization of the error signal in the image encoding block 120 is based on the quantization parameter QP determined according to the data fullness state information provided from the output side transmission buffer 140 described later through the line L21. Is adjusted.

Meanwhile, in the quantization step of the image encoding block 120, as described above, the process of quantizing the video signal with a step size according to the quantization parameter QP determined according to the data fullness state information of the output side transmission buffer 140 is discrete. The DCT transform coefficient of the cosine transformed M × N block is divided by a value of 2 * QP. The quantization parameter (QP) has different values for each macro block (16x16 block) according to the data fullness of the transmission buffer 140. It can also have an integer value from 1 to 31. Therefore, in the image encoding block 120, when the data-filled state of the transmission buffer 140 rises above the predetermined value, the QP value is increased to reduce the amount of bits generated after entropy encoding, and conversely, the data-filled state of the transmission buffer 140 is increased. When is lowered below a predetermined value, the QP value is lowered to reduce the amount of bits generated after entropy encoding.

In addition, although the detailed illustration in FIG. 1 is omitted, in the quantizer included in the image encoding block 120, the average quantization parameter (AQP) for every frame provided from the weight generation block 220 described later according to the present invention. The weights based on Δ are used to determine the elimination (ie, filtering) of high frequency components that are relatively insensitive to human vision in the quantization step. That is, the quantizer in the image coding block 120 adaptively limits the bandwidth limited in the low pass filtering in the quantization step. Therefore, the present invention can appropriately adjust the quantization step size that causes deterioration of image quality in the reproduced video even in a complex video accompanied by an increase in the encoded bit generation amount. As described above, a detailed process of adaptively limiting the bandwidth in the low pass filtering in the quantization step by using a weight set based on the average quantization parameter (AQP) of each frame unit will be described later in detail.

Next, the quantized DCT transform coefficients on line L13 are sent to entropy coding block 130 and image decoding block 150, respectively. Here, the quantized DCT transform coefficients provided to the entropy coding block 130 are encoded, for example, through a variable length coding scheme, and provided to the transmission buffer 140 on the output side. It is delivered to the transmitter not shown for the transmission of.

On the other hand, the quantized DCT transform coefficients on the line L13 provided from the image coding block 120 to the image decoding block 150 are converted into a reconstructed frame signal through inverse quantization and inverse discrete cosine transform, and then added to the adder 160. The adder 160 adds the predicted current frame signal provided from the current frame prediction block 180 described later through the reconstructed frame signaling line L19 from the image decoding block 150 to reconstruct the previous image. A frame signal is generated, and the previous frame signal reconstructed as described above is stored in the second frame memory 170. Therefore, the immediately previous frame signal for every frame encoded through such a path is continuously updated, and the reconstructed previous frame signal thus updated is provided to the current frame prediction block 180 described later for motion estimation and compensation. do.

On the other hand, in the current frame prediction block 180, the current frame signal on the line L11 provided from the first frame memory 100 described above and the reconstructed previous frame on the line L15 provided from the second frame memory 170 described above. Based on the signal, the current frame is predicted in units of a predetermined block (eg, 8 × 8 DCT block) from a preset search range (eg, 16 × 16 search range) of the previous frame reconstructed using a block matching algorithm. The predicted current frame signal is generated and provided to the subtractor 110 and the adder 160, respectively. At this time, the switch SW on the line L19 is turned on / off according to the control signal CS from the system controller (not shown). When the switch SW is on, it means that the current encoding mode is the inter mode. On the contrary, when off, this means that the current encoding mode is an intra mode. Accordingly, the subtractor 110 provides an error signal between the current frame signaling prediction frame signals to the image encoding block 120 during inter-mode encoding, and transmits the current frame signal itself to the image encoding block 120 during intra-mode encoding. to provide.

The current frame prediction block 180 also generates a set of motion vectors for each selected block (8x8 block) on line L17 and provides it to the entropy coding block 130 described above. Here, the sets of motion vectors detected are the displacements between the most similar block predicted in the block of the current frame (8x8 block) and the preset search area (e.g., 16x16 search range) in the previous frame. Accordingly, in the entropy coding block 130 described above, the quantized DCT transform coefficients on the line L13 together with the sets of the motion vectors on the line L17 generate an encoded bit stream by encoding, for example, through a variable length coding technique. .

Meanwhile, according to the present invention, the quantization parameter value QP determined in units of macroblocks in the quantizer (not shown) in the image encoding block 120 according to the data full state of the output side transmission buffer 140 is quantized through the line L23. The parameter QP is input to the average value calculation block 210.

Accordingly, the QP average value calculation block 210 calculates a QP average value, that is, an AQP value, for every frame. That is, in one QP average value calculation block 210, after encoding (quantization, DCT, and entropy encoding) of image data of a frame, an AQP value is calculated by averaging QP values for all macroblocks of the encoded frame. This means that the AQP value of the currently encoded frame is used to calculate the complexity in the encoding process for the next frame by using the characteristics of the video signal that the frames adjacent to each other in time are not largely changed. As a result, when the current encoded image is a complex image, the AQP value calculated in the QP average value calculation block 210 will have a large value, and when the simple image is relatively easy to encode, the AQP value is a small value. Will be.

More specifically, if the video signal of one frame to be encoded currently has the size of MxN and each macroblock has the size of LxL, the number of macroblocks of this video signal is (ML) x (NL). Therefore, the number of QPs generated after the video signal of one frame is encoded is (ML) x (NL). For example, assuming 16x16 macroblocks for a 352x288 input video signal, the number of macroblocks is (352/16) x (288/16), which corresponds to 22x18, that is, 396. Therefore, in the present invention, the AQP value calculated by averaging the QPs of all such macroblocks is used as the complexity of the next video signal input for encoding. In this case, the calculated AQP value may be considered to be related to the information amount (bit generation amount) of the image. If the current encoded image is complex, the calculated AQP value will be relatively large, and vice versa. The value will be smaller. Then, the AQP value calculated in the QP average value calculation block 210 is provided to the weight generation block 220 of the next stage.

Meanwhile, the weight generation block 220 limits the pass band of the frequency domain in the quantization step of the image coding block 120 based on the AQP value provided from the QP average value calculation block 210, that is, to the human eye. A weight is generated to selectively limit the high frequency components of the relatively insensitive image. That is, the weight generation block 220 may be a plurality of preset weights (eg, four weights for setting different filter coefficients) for limiting bandwidth during low pass filtering before the quantization step based on the calculated AQP value. The weight corresponding thereto is determined and generated on the line L25 and provided to the image encoding block 120. Here, the weights determined and provided to the image coding block 120 divide the frequency values according to the B values necessary to distinguish the frequency domains calculated as follows immediately before quantization of the input image (ie, the DCT transform coefficient). In this case, the area division value B necessary for dividing the frequency domain is calculated by the following method, and a process of setting the frequency domain using the area division value B will be described later in detail with reference to FIG. will be.

In more detail, the process of calculating the frequency domain division value B using the AQP value output from the QP average value calculation block 210 in the weight generation block 220 is as shown in Equation (1) below.

In the above formula (1), ROUND is an operation that takes an integer value by rounding. Therefore, as a result of Equation (1), in the conventional encoder having an integer value between 1 and 31, the B value for frequency domain division has an integer value between 1 and 4. That is, when the calculated AQP value is 31, since the ROUND {(31-16) / 5} is 3, the frequency domain division value B becomes 4. Therefore, according to the value B calculated in this way, as shown in FIG. 2 as an example, the frequency range to be limited is determined. This operation is to automatically adjust the bandwidth limiting in the two-dimensional low pass filtering before the quantization step in the image coding block 120 according to the AQP value.

Next, a process of adjusting weights for frequency band limitation according to the present invention using the B value calculated as described above will be described in detail with reference to FIG. 2 attached as an example.

Referring to FIG. 2, when the B value calculated by the weight generation block 220 is 1, the image encoding block 120 may perform discrete cosine transform on the DCT transform coefficients obtained by discrete cosine transforming an error signal input from the subtractor 110. Quantization is performed directly to a predetermined step size determined based on the data fullness state information of the transmission buffer 140 without giving weight for frequency band limitation, that is, without passing through low pass filtering. This means that the input image is not complicated and simple, and is encoded (quantized) as it is without low pass filtering as in the conventional method.

Unlike the above, as illustrated in FIG. 2, when the B value has a value of 2, 3 or 4, the image coding block 120 divides all frequencies below the dotted line corresponding to each value by 2. Perform quantization. That is, when the B value is 4, frequencies below the dotted line such as F (1,7), F (2,6), etc. in FIG. 2 are all divided by 2 to be encoded. Therefore, when the encoding (quantization) is performed through the above process according to the present invention, the low frequency signal, which is a visually important component, can be encoded while generating less quantization error. If the weight is not limited to the low pass bandwidth of the frequency as in the present invention despite the complicated image, as a result, the amount of bits generated after encoding increases and the quantization step size becomes large, so that all frequency bands (high frequency to low frequency) Quantization error is generated a lot, resulting in deterioration of image quality due to quantization in the playback image of the receiving side.

Accordingly, in the weight generation block 220 of the present invention, the complexity of the image to be currently encoded is calculated by using the AQP value calculated by averaging the respective QPs generated in the immediately encoded frame, and based on the complexity calculation result. By generating weights to limit the bandwidth of the high-frequency components of the image, the low-pass filtering before quantization (ie, two-dimensional low-pass filtering) reduces the bit rate of the image first, so even in complex images, While suppressing an excessive increase in the step size, it is possible to sufficiently encode (quantize) low frequency components that are visually important information. Therefore, it is possible to suppress the deterioration of the image quality due to the quantization error in the reproduced video reconstructed by the receiving side decoding system.

Next, the process of performing conventional DCT and quantization and the process of quantizing while weighting based on the complexity of the image according to the present invention will be described based on the results of experiments performed by the present inventor on an image signal of one block as an example. do.

In the following description, an example of reproduction of a video signal of a block composed of 8x8 pixels is performed through conventional DCT, quantization, inverse quantization, and IDCT processes. In this case, it is assumed that the DCT conversion coefficient is divided by the step size QP * 2, and the step size is 2 (QP = 1).

Table 1 below shows a video signal for one block (8x8 pixels). As can be seen from Table 1, it can be seen that each pixel has a level value between 0 and 255. Represents each pixel value for the position of.

Table 2 below shows the result of performing 8x8 DCT in the horizontal and vertical directions on a block of image signals having pixel values as shown in Table 1 above. It means the frequency component in the direction.

Table 3 below shows the result of performing quantization on a DCT transform coefficient block 8x8 having frequency components as shown in Table 2 above. The quantization step size in this process was calculated as QP * 2, taking the case where QP is 1, that is, the step size is 2. As a result, the value at each position corresponds to the result of dividing the value of Table 2 above by 2. Therefore, as a result of performing this operation (quantization), each of the quantized DCT transform coefficients is smaller in size than the respective DCT transform coefficients shown in Table 2, and the number having a value of 0 increases. Able to know. As a result, this means that the amount of information to be transmitted after encoding is reduced. If the QP value at this time becomes large, the resulting quantized DCT transform coefficients in Table 3 are not only smaller in size but also smaller in number with a value of 0, so that after encoding The amount of data to be transmitted will be further reduced. As described above, this is a conventional method of controlling the amount of information to be transmitted by adjusting the QP value in the quantization step.

Meanwhile, in the present invention, not only the bit amount generated after encoding is adjusted through the conventional quantization step size adjustment as described above, but also each macroblock of a frame that has been previously encoded (quantized, DCT, and entropy encoded). Using the AQP value obtained by averaging the values of the QPs, the complexity of the image to be currently encoded is calculated, and based on the calculated complexity, the image to be encoded is a complex image (that is, the calculated AQP value is a relatively large value). Image), a plurality of weights having a predetermined high frequency cutoff level are adaptively given before the quantization step to limit the pass frequency band of the DCT transform coefficients.

That is, Table 4 shows, as an example, DCT conversion given a weight corresponding to the case where the area division value B calculated in the weight generation block 220 shown in FIG. 1 is 4 for one block (8x8 DCT block). Shows the result of quantizing the coefficients. The result of Table 4 shown above is a result of quantizing the frequency part below the dotted line by 2 in Table 4 as compared with the quantized DCT transform coefficient values of Table 3, and has a large number having a level of 0. Also, it can be seen that the magnitude of each quantized DCT transform coefficient value is also two times smaller. As a result, in the present invention, by applying a weight based on the complexity of the image determined based on the AQP value of the previous frame, adaptively removing high frequency components of the image and performing quantization, an excessive step size in the quantization step is achieved. It is possible to effectively adjust the amount of bits generated after encoding without increasing. The dotted line shown in Table 4, which is determined according to the B value of 4, corresponds to the portion of the weighted B value 4 in FIG.

As described above, according to the present invention, the complexity of an image to be currently encoded is calculated by using an AQP value obtained by averaging QP values generated for each macroblock at the time of quantization in a frame immediately encoded, and calculating the calculated value. According to the result, if the current input image is a complex image, the corresponding weight is applied to remove the high frequency components of the image insensitive to the human's vision, and then the DCT, quantization, etc. encoding is performed, thereby causing excessive steps in the quantization step. The amount of bits generated after encoding can be effectively adjusted without increasing the size. Therefore, according to the present invention, it is possible to effectively reduce image quality deterioration due to quantization error inevitably present in a reproduced image when reconstructing and displaying an encoded image.

Claims (4)

The differential signal between the input current frame and the prediction frame obtained through motion estimation and compensation using the current frame and the reconstructed previous frame is encoded by compression encoding by encoding means including discrete cosine transform, quantization, and entropy encoding. A video encoding system having a bit generation amount adjusting function for generating a bit stream, and adding quantization whose step size is adjusted based on a quantization parameter determined according to the full state information of the bit stream stored in an output buffer. Means for calculating an average quantization parameter value of the previous frame by averaging each quantization parameter value generated for each macroblock of the immediately preceding frame; and encoding the average quantization parameter value calculated in the previous frame.In the low pass filtering of each DCT transform coefficient block generated through the discrete cosine transform, referred to as the complexity of the frame to be currently encoded through the stage, the calculated weight among the plurality of preset weights that adaptively limits the bandwidth Weighting means for determining a weight corresponding to an average quantization parameter value of a previous frame and providing the weighting means to the encoding means, wherein the encoding means is based on the determined weight before quantization of the respective DCT transform coefficient blocks. And restricting a high frequency component of each DCT transform coefficient block by limiting a pass frequency band of each DCT transform coefficient block. The image encoding system of claim 1, wherein the preset plurality of weights comprises four weights of integer values for setting different filter coefficients of the respective DCT transform coefficient blocks. . The bit generation amount adjusting function of claim 1 or 2, wherein the encoding means selectively filters high frequency components of the respective DCT transform coefficient blocks according to the determined weight through two-dimensional low pass filtering. Image coding system having a. The bit generation amount according to claim 3, wherein the encoding means performs quantization by dividing harmonic components below the determined weight by 2 to selectively filter high frequency components of the respective DCT transform coefficient blocks. Image coding system with control function.