JP6398149B2 - Moving picture coding apparatus and program, and moving picture coding system - Google Patents

Description

  The present invention relates to a moving image coding apparatus, a program, and a moving image coding system, and for example, performs coding and decoding of a moving image based on a Distributed Video Coding (hereinafter referred to as DVC) system. Applicable when doing.

  The DVC method is a moving image encoding method that performs encoding and decoding of a moving image based on the Slepian-Wolf theory or the Wyner-Ziv theory (see Non-Patent Document 1).

  In the DVC method, a code (hereinafter referred to as a Wyner-Ziv code) for reconstructing a coding target image from a prediction image (hereinafter referred to as a decoder predicted image) of a coding target image generated by a decoder is directly used as the decoder predicted image. It is characterized by generating without reference. Due to this feature, the DVC encoding apparatus does not need to include a complicated prediction image generation unit, and the amount of calculation related to encoding can be reduced.

  The problem of the DVC method is that it is difficult to obtain a Wyner-Ziv code rate (code amount) necessary and sufficient to generate a decoded image from a decoder predicted image. If there are too many Wyner-Ziv codes, the compression effect cannot be obtained, and if there are too few Wyner-Ziv codes, a decoded picture cannot be generated from the predicted decoder picture.

  The technology described in Non-Patent Document 1 is a method of searching for a necessary and sufficient rate by continuing to request (continue feedback) an additional Wyner-Ziv code until a decoded image is generated from a decoder predicted image (hereinafter referred to as decoder rate control). This method is used to solve the rate problem.

  However, the decoder rate control method has a disadvantage that the use is limited because it cannot be encoded without the presence of a decoding process via a network.

  In order to eliminate this shortcoming, in Non-Patent Document 2 and Non-Patent Document 3, an encoder-predicted image (a predictive image that can be generated without requiring a calculation as complicated as a decoder-predicted image) is generated by an encoder, and is encoded. A method for controlling a rate by comparing an image with an encoder predicted image (hereinafter referred to as an encoder rate control method) has been proposed.

  FIG. 7 is a block diagram showing a schematic configuration of a conventional video encoding apparatus based on the encoder rate control method.

  The moving image encoding device 31 illustrated in FIG. 7 receives an encoding target image (non-key frame image) PIC1 and a reference image PIC2, and outputs a bit stream BS.

  The moving image encoding device 31 estimates a necessary and sufficient rate for the decoder to decode from the encoding target image PIC1 and the reference image PIC2, and uses the rate control unit 302 that outputs the rate S302, and the rate S302. A Wyner-Ziv encoding unit 103 that performs Wyner-Ziv encoding on the encoding target image PIC1.

  For example, as in the technique described in Non-Patent Document 2, the rate control unit 302 generates an encoder predicted image from the reference image PIC2, calculates an error statistic between the encoding target image PIC1 and the encoder predicted image, From this statistic, a rate S302 necessary and sufficient for the decoder to decode is estimated. Further, for example, the rate control unit 302 generates an encoder predicted image from the reference image PIC2 as in the technique described in Non-Patent Document 3, counts the number of bit inversions between the encoding target image PIC1 and the encoder predicted image, and performs bit inversion. Based on the bit inversion probability that can be calculated from the number, a rate S302 necessary and sufficient for the decoder to decode is estimated.

  Although omitted in FIG. 7, not only the bit stream obtained by encoding the encoding target image PIC1 but also the encoding target image PIC1 from the moving image encoding device 31 to the opposing moving image decoding device (not shown). Compressed data obtained by encoding a key frame image having a different time with an encoding method different from the encoding method of the encoding target image PIC1 is also transferred.

  The reference image PIC2 is, for example, an original image or a decoded image of a key frame image captured at the immediately preceding time or immediately following time, an image obtained by performing image processing on these images, and the like. Key frame images are often H.264 images. It is compressed by an existing encoding technique such as H.264 / AVC (however, the key frame image encoding method is not limited).

B. Girod, A.M. Aaron, S.A. Rane, and D.D. Rebolo-Monedero, “Distributed video coding,” Proceedings of the IEEE, Vol. 93, pp. 71-83, Jan. , 2005 C. Brites, F.M. Pereira, “Encoder rate control for domain domain Wyner-Ziv video coding,” ICIP 2007, USA, September, 2007 K. Sakomizu, T .; Yamazaki, S .; Nakagawa and T.A. Nishi, “A real-time system of distributed video coding,” Proc. of Picture Coding Symposium (PCS'10), pp. 538-541, Nagaya Japan, Nov. , 2010 M.M. Tagliasacchi A. Majmardar and K.M. Ramchandran, “A distributed-source-coding based robust-spatial-temporal scalable video codec,” Proc. Picture Coding Symposium, Citeseeer, 2004 K Sakomizu, T .; Nishi and T. Onee, “A hierarchical motion smoothing for distributed scalable video coding,” Proc. of Picture Coding Symposium (PCS'12), pp. 209-212, Krakow, Poland, May. 2012.

  In a moving picture coding apparatus based on the DVC system including a rate control unit, it is difficult to estimate a rate necessary and sufficient for decoding by a decoder without using a predicted decoder picture.

  As described above, conventionally, information on the difference between the encoding target image and the encoder predicted image is calculated, and the rate is estimated from the information on the difference.

  However, the information regarding the difference between the encoding target image and the encoder predicted image is not equal to the information regarding the difference between the encoding target image and the decoder predicted image. This error adversely affects the rate estimation accuracy, and as a result, the coding efficiency is deteriorated.

  The deterioration of the encoding efficiency can be solved by increasing the load on the encoder, for example, by increasing the prediction accuracy of the encoder predicted image. However, in this solution, the feature of the DVC method that the encoder is lightweight cannot be used.

  Therefore, there is a demand for a moving picture coding apparatus and program, and a moving picture coding system that can improve the coding efficiency by improving the accuracy of rate estimation while maintaining the lightness of processing on the encoder side. .

According to a first aspect of the present invention, there is provided a moving image encoding apparatus for compressing a moving image based on a distributed video encoding method, wherein (1) an encoding target image that is a non- key frame and a reference image that is a key frame are encoded as described above. Translational motion region identification means for identifying a translational motion region in the target image; and (2) a rate that defines a code amount necessary for the moving image decoding device to generate a decoded image, and information on the translational motion region . Different weights are applied to the information in the non-translational motion region , which is a region other than the translational motion region , and at least weighting is performed so that the information amount of the translational motion region is small . using information of the information and the non-translational movement region of the translation region, and determining the rate control means, (3) applying the determined rate, marks for encoding the encoding target image Characterized in that it comprises a means.

The moving image encoding program of the second aspect of the present invention provides a computer mounted on a moving image encoding apparatus that compresses a moving image based on a distributed video encoding method, and (1) an encoding target image that is a non-key frame and A translational motion region identifying means for identifying a translational motion region in the encoding target image from a reference image that is a key frame ; and (2) a code amount necessary and sufficient for the video decoding device to generate a decoded image. Different rates are applied to the information of the translational motion region and the information of the non- translational motion region other than the translational motion region , and at least the information amount of the translational motion region is reduced. performs weighting such as by using the information of the information and the non-translational movement region of the translation region was weighted, and rate control means for determining a rate determined (3) And use, characterized in that to function as an encoding means for encoding the encoding target image.

  According to a third aspect of the present invention, there is provided a moving image code comprising: a moving image encoding device that compresses a moving image based on a distributed video encoding method; and a moving image decoding device that decodes encoded data from the moving image encoding device. In the encoding system, the moving image encoding device according to the first aspect of the present invention is applied as the moving image encoding device.

  According to the present invention, it is possible to provide a moving picture coding apparatus and program, and a moving picture coding system capable of improving rate estimation accuracy and improving coding efficiency while maintaining lightness of processing on the encoder side.

It is a block diagram which shows the structure of the moving image encoding system of 1st Embodiment, and about the moving image encoding apparatus of 1st Embodiment, the functional detailed structure inside is also shown. It is a flowchart which shows operation | movement of the moving image encoder of 1st Embodiment. It is explanatory drawing (the 1) of operation | movement of the translation area | region identification part in the moving image encoder of 1st Embodiment. It is explanatory drawing (the 2) of operation | movement of the translation area | region identification part in the moving image encoder of 1st Embodiment. It is explanatory drawing which shows the detailed structure of the integration part in the moving image encoder of 2nd Embodiment. It is a flowchart which shows the detailed operation | movement of the integration part in the moving image encoder of 2nd Embodiment. It is a block diagram which shows schematic structure of the conventional moving image encoder based on an encoder rate control system.

(A) First Embodiment Hereinafter, a first embodiment of a moving image encoding apparatus and program and a moving image encoding system according to the present invention will be described with reference to the drawings.

(A-1) Configuration of the First Embodiment FIG. 1 is a block diagram showing the configuration of the video encoding system of the first embodiment. Regarding the video encoding device of the first embodiment, The internal detailed structure is also shown.

  In FIG. 1, a moving image encoding system 1 according to the first embodiment includes a moving image encoding device (hereinafter referred to as an encoder) according to the first embodiment that encodes an encoding target image (non-key frame image) PIC1. 11, a key frame encoding device 12 that encodes the key frame image PIC 3, and a moving image decoding that decodes the encoded data (bit stream) BS from the moving image encoding device 11 A device 21 (hereinafter also referred to as a decoder) 21 and a key frame decoding device 22 for decoding encoded data from the key frame encoding device 12 are included.

  Since the key frame encoding device 12, the moving image decoding device 21, and the key frame decoding device 22 are the same as those in the conventional moving image encoding system according to the DVC method, detailed description thereof is omitted.

  The moving image encoding device 11 of the first embodiment has a detailed configuration as shown in FIG. The video encoding device 11 may be constructed by connecting various circuits in hardware, and a general-purpose device having a CPU, a ROM, a RAM, and the like executes a video encoding program to execute a moving image. It may be constructed so as to realize a function as an image encoding device. Regardless of which construction method is applied, the detailed functional configuration of the moving image encoding apparatus 100 is the configuration shown in FIG.

  The moving image encoding device 11 includes a translational motion region identification unit 101, a rate control unit 102, and a Wyner-Ziv encoding unit 103.

  The translational motion region identification unit 101 is a region that is moving at the same speed in the same direction in the encoding target image PIC1 from the input encoding target image PIC1 and the reference image PIC2 (referred to as a translation region as appropriate). In addition, a translational motion region in which the movement is a region associated with the movement of the moving body (referred to as a motion region as appropriate) is obtained, and a translational motion region indicating whether the motion is a translational motion region or a non-translational motion region The identification signal S101 is output to the rate control unit 102.

  The rate control unit 102 estimates a rate (code amount) necessary and sufficient for the video decoding device 21 to decode based on the input encoding target image PIC1, the reference image PIC2, and the translational motion region identification signal S101. Then, the rate S102 obtained by the estimation is output to the Wyner-Ziv encoding unit 103.

  The Wyner-Ziv encoding unit 103 performs Wyner-Ziv encoding so that the input encoding target image PIC1 satisfies the code amount defined by the input rate S102, and the obtained encoded data The bit stream BS is transmitted to the video decoding device 21. The Wyner-Ziv encoding unit 103 uses information related to key frame images such as encoded data of key frame images and key frame images when encoding the encoding target image PIC1.

  The translational motion region identification unit 101 includes an encoding target image reduction unit 104, a reference image reduction unit 105, a motion vector estimation unit 106, a motion region identification unit 107, a translational region motion vector calculation unit 108, and a translational region identification. Unit 109 and integration unit 110.

  The encoding target image reduction unit 104 reduces the input encoding target image PIC1 and outputs the reduced encoding target image S104 to the motion vector estimation unit 106 and the motion region identification unit 107.

  The reference image reduction unit 105 reduces the input reference image PIC2 and outputs the reduced reference image S105 to the motion vector estimation unit 106 and the motion region identification unit 107.

  The motion vector estimation unit 106 estimates the motion vector S106 from the reduction encoding target image S104 given from the encoding target image reduction unit 104 and the reduction reference image S105 given from the reference image reduction unit 105, and obtains it. The obtained motion vector S106 is output to the motion region identification unit 107, the translation region motion vector calculation unit 108, and the translation region identification unit 109.

  The motion region identification unit 107 is supplied from the reduction encoding target image S104 provided from the encoding target image reduction unit 104, the reduced reference image S105 provided from the reference image reduction unit 105, and the motion vector estimation unit 106. From the motion vector S106, a motion region is identified, and a motion region identification signal 107 that defines a motion region or a non-motion region is output to the integration unit 110.

  The translation region motion vector calculation unit 108 estimates the motion vector of the translation region from the motion vector S106 given from the motion vector estimation unit 106, and outputs the obtained translation region motion vector S108 to the translation region identification unit 109.

  The translation region identification unit 109 defines a translation region or a non-translation region based on the motion vector S106 given from the motion vector estimation unit 106 and the translation region motion vector S108 given from the translation region motion vector calculation unit 108. A region identification signal 109 is generated and output to the integration unit 110. For example, the translation region identification unit 109 generates the translation region identification signal S109 by comparing the magnitude of the difference between the motion vector S106 and the translation region motion vector S108 with a threshold value.

  The integration unit 110 identifies a translational motion region from the motion region identification signal S107 and the translational region identification signal S109, generates a translational motion region identification signal S101, and outputs the translational motion region identification signal S101 to the rate control unit 102. For example, the motion region identification signal S107 is a signal that takes logic “1” in the motion region and logic “0” in the non-motion region, and the translation region identification signal S109 has logic “1” in the translation region and logic “1” in the non-translation region. If the signal is “0”, the integration unit 110 generates a translational motion region identification signal S101 by calculating a logical product of the two identification signals S107 and S109.

(A-2) Operation of First Embodiment Next, the operation of the moving image encoding device 11 of the first embodiment will be described with reference to the drawings.

  Note that the operations of the key frame encoding device 12, the moving image decoding device 21, and the key frame decoding device 22 are the same as those in the prior art, and a description thereof will be omitted.

  FIG. 2 is a flowchart illustrating the operation of the moving image encoding device 11 according to the first embodiment.

  When a new encoding target image PIC1 becomes a processing target, the encoding target image reduction unit 104 generates a reduction encoding target image S104 from the encoding target image PIC1, and the reference image reduction unit 105 performs reference. A reduced reference image S105 is generated from the image PIC2 (step ST101). Any existing algorithm may be applied as a reduced image generation algorithm. For example, a reduction method using a Bilinear function or a Bicubic function can be applied. Further, the reduction encoding target image S104 and the reduction reference image S105 may be subjected to image processing. For example, in the case of the DVC method having a scalable structure as described in Non-Patent Document 4 and Non-Patent Document 5, the reduced image of the encoding target image PIC1 and the reduced image of the reference image PIC2 are extracted from the information of the reduced image layer. Thus, a reduced image may be generated.

  In the motion vector estimation unit 106, a motion vector S106 for each block is generated from the reduced encoding target image PIC104 and the reduced reference image PIC105 (step ST1O2). The motion vector can be obtained, for example, by searching for a motion vector using a block matching method or a gradient method. Here, the block size may be arbitrarily determined.

  The motion region identification unit 107 generates a motion region identification signal S107 for each block from at least one of the reduction encoding target image S104, the reduced reference image S105, and the motion vector S106 (step ST1O3).

  As a method for discriminating between the motion region and the non-motion region, for example, any of the following methods M1 to M5 can be applied.

  (M1) Referring to the motion vector S106, if the motion vector S106 is larger than a predetermined threshold, it is determined to be a motion region, and other blocks are determined to be non-motion regions.

  (M2) Referring to the reduction encoding target image S104 and the reduction reference image S105, the magnitude of the difference between the reduction encoding target image S104 and the reduction reference image S105 is calculated using Sum of Absolute Difference (SAD) or the difference absolute value sum. A block is evaluated for each block using an index called mean absolute difference (MAD), and a block whose difference is larger than a predetermined threshold is determined to be a motion region. The block is determined as a non-motion area.

  (M3) The absolute value of the difference between the reduced encoding target image S104 and the reduced reference image S105 is obtained for each pixel with reference to the reduced encoding target image S104 and the reduced reference image S105, and the absolute value of the difference is determined in advance. The number of pixels larger than the set threshold value is counted for each block, and a block in which the number of pixels is larger than a predetermined threshold value is determined as a motion region, and other blocks are determined as a non-motion region.

  (M4) This is a method for referring to the reduced encoding target image S104 and the reduced reference image S105. For the reduced encoding target image S104, an average value of pixel values in the first average calculation block is obtained for each block having a size for average value calculation (referred to as a first average calculation block), By subtracting the average value obtained from each pixel value included in the first average calculation block, an AC component image of the reduction encoding target image is obtained. The size of the first average calculation block may be the frame size of the reduction encoding target image (this is the largest size in this case), or may be a size obtained by dividing the frame size into a plurality of sizes. Similarly, for the reduced reference image S105, the average value of the pixel values in the second average calculation block for each block (referred to as a second average calculation block) having an average value calculation size. And an AC component image of the reduced reference image is obtained by subtracting the average value obtained from each pixel value included in the second average calculation block. The size of the second average calculation block may be the frame size of the reduced reference image (this is the largest size in this case), or may be the same size as the size of the first average calculation block. . Then, the absolute value of the difference between the AC component image of the reduction encoding target image and the AC component image of the reduced reference image is obtained for each pixel, and the absolute value of the difference for each pixel is greater than a predetermined threshold value. Are counted for each block having a size for region identification, and a block having this number of pixels larger than a predetermined threshold is determined as a motion region, and other blocks are determined as a non-motion region.

  (M5) The method M1 and one or more of the methods M2 to M4 are combined to identify the motion region or the non-motion region. The region determined as the motion region by the method M1 is determined as the motion region, and the region determined as the non-motion region by the method M1 is finally used by using one or more of the methods M2 to M4. Determine whether the region or non-motion region. The method M5 is a method for preventing the calculation amount of the methods M2 to M4 from increasing.

  Here, in the first embodiment, all processes for identifying the translational motion region are performed on the reduced images (reduced encoding target image S104 and reduced reference image S105), thereby increasing the amount of calculation. I try to suppress it.

  However, by searching for a motion vector for a reduced image, the detection accuracy of the motion vector is deteriorated in proportion to the reduction magnification, and it is difficult to determine the motion region. For example, a region where only one pixel (pixel) is moving is a region that should be determined as a motion region, but when a motion vector search is performed on a reduced image, the size of the estimated motion vector There is a high possibility that it will be zero. By applying the method M2 to the method M5, it is possible to carry out the determination of the motion region with high accuracy within the constraint of identifying the low-computation translational motion region using the reduced image.

  In order to identify the translation region from the motion vector S106 in the translation region motion vector calculation unit 108 in parallel with or before or after the generation of the motion region identification signal S107 for each block in the motion region identification unit 107 described above. The translation region motion vector S108 is generated (step ST104).

  Here, the translation region motion vector S108 is, for example, an average motion vector in the frame of the motion vector S106, or a most frequent motion vector in the frame. In the former case, the average of the motion vectors S106 in the frame is the translation region motion vector S108. In the latter case, the motion vector that appears most frequently among the motion vectors S106 in the frame is the translation region motion vector S108.

  When the translation region motion vector S108 is obtained, the translation region identification unit 109 generates a translation region identification signal 109 from the motion vector S106 and the translation region motion vector S108 (step ST1O5).

  The translation area identification signal S109, for example, subtracts the translation area motion vector S108 from the motion vector S106 of each block (each area). If the magnitude of the difference information is less than a predetermined threshold value, the translation area identification signal S109 is determined to be a translation area. If the size of the information is equal to or greater than the above-described threshold, the information is generated by determining the non-translational region. This method is based on the assumption that the translation region motion vector S108 is close to the motion vector of the translation region. All the pixels included in the translation region have substantially the same motion vector. For this reason, the motion vector of the translation region tends to have a large influence on the translation region motion vector S108. In other words, there are many cases where the above hypothesis is met. By searching for a region having a motion vector close to the translation region motion vector S108, the translation region having the greatest influence can be obtained.

  Thereafter, in the integration unit 110, a translational motion region identification signal S101 is generated from the motion region identification signal S107 and the translational region identification signal S109 (step ST1O6).

  Specifically, it is determined as a translational motion region only when the motion region identification signal S107 is determined as a motion region and the translational region identification signal S109 is determined as a translational region. By determining the translation motion region, a translation motion region identification signal S101 is generated.

  Hereinafter, the operation of the translational motion region identification unit 101 will be described with reference to FIGS. 3 and 4.

  3A shows the reference image PIC2, and FIG. 3B shows the encoding target image PIC1. In the case of FIG. 3, it is assumed that the reference image PIC2 is an image of a frame immediately before the encoding target image PIC1.

  A pedestrian, a house, and a cloud are imaged in the encoding target image PIC1 and the reference image PIC2. A pedestrian is a moving object, a house is a stationary object, and a cloud moves very slowly and can be regarded as a stationary object. An imaging device (imaging camera or the like) that captures an image moves in a direction opposite to that of a pedestrian. For this reason, houses, clouds, and other backgrounds that are stationary objects are moving from the right to the left of the image. On the other hand, the pedestrian is walking from the left to the right of the image.

  In the case of such a moving image, a house, a cloud, and other background objects are translated from right to left. On the other hand, pedestrians often perform complex exercises such as waving their hands and moving their feet.

  Therefore, it is determined that the object region of the house, the cloud, or other background is a translation region ("A" is given in FIG. 3C), and the translation region identification signal S109 for these regions indicates the translation region. Takes a logical value to represent. On the other hand, it is determined that a region including a pedestrian performing complex exercise is a non-translation region ("B" is given in FIG. 3C), and the translation region identification signal S109 for these regions is non-translational. Takes a logical value representing the translation region.

  FIG. 4A shows the motion vector S106 (outline) obtained for the image of FIG. Most of the areas that do not include pedestrians have a leftward translational motion that accompanies the panning of the imaging device, have the same leftward motion vector, and the areas that are surrounded by dotted lines that include pedestrians are pedestrians. The motion vector has a downward or rightward motion vector corresponding to the complicated motion of the.

  When the average motion vector in the frame of the motion vector S106 is applied as the translation region motion vector S108, most of the motion vectors in the region are leftward. Therefore, when the translation region motion vector S108 is obtained, the upper part of FIG. The translation region motion vector S108 substantially leftward as described in (1) is obtained. A translation region in which the translation region motion vector S108 is subtracted from each motion vector S106 and the magnitude of the difference is less than the threshold is included in the area to which “A” is given in FIG. 4B, and the magnitude of the difference is the threshold. The above non-translational regions are included in the area given “B” in FIG.

  Also, when the most frequent motion vector in the frame of the motion vector S106 is applied as the translation region motion vector S108, the motion vector in the most part is leftward, so that the leftward motion vector is used as the translation region motion vector S108. In this case as well, the translation region is included in the area given “A” in FIG. 4 (B), and the non-translation region is in the area given “B” in FIG. 4 (B). include.

  As described above, the translational region identification signal S109 that divides the image into the translational region and the non-translational region can be generated by the operation of the translational motion region identification unit 101.

  Here, the translation area identifying unit 109 only identifies the translation area, and the translation area includes a stationary area. This is because the stationary state of the object is also one translation, such as when the imaging device pans.

  The still region is a region with high prediction accuracy not only in the decoder predicted image but also in the encoder predicted image. The fact that there is an error or bit inversion with respect to the static region despite the fact that the prediction accuracy is high originally suggests that this is an error or bit inversion that cannot be compensated in the first place. Such an error or bit inversion occurs due to, for example, flickering of illumination or noise of the imaging device. Illumination flicker, imaging device noise, and the like are random changes that are extremely difficult to compensate by prediction.

  Unlike the first embodiment, when only the translation region identification is applied and the motion region identification is not applied, the above-described error occurring in the stationary region is detected at the rate estimation described later. There is a risk of underestimation.

  In order to avoid such an underestimation, the motion region identification unit 107 is provided. The motion region identification unit 107 determines whether each region is a motion region (non-stationary region) or a non-motion region (stationary region), and the integration unit 110 obtains a translational and moving region. Thus, a translational motion region that can avoid the above-described underestimation is obtained.

  The translational motion region identification unit 101 only needs to be able to recognize the translational motion region, and is based on a combination of the motion vector estimation unit 106, the motion region identification unit 107, the translational region motion vector calculation unit 108, the translational region identification unit 109, and the integration unit 110. The recognition of the translational motion region in the first embodiment is an example for recognizing the translational motion region with a low calculation amount.

  When the translational motion region identification signal S101 is generated, the rate control unit 102 uses the rate target bit rate BS to be transmitted to the video decoding device 21 from the encoding target image PIC1, the reference image PIC2, and the translational motion region identification signal S101. (Code data amount) S102 is estimated (step ST107).

  The rate control unit 102 compares the encoding target image PIC1 with the encoder predicted image generated in the rate control unit 102 while weighting the translational motion region and the non-translational motion region, so that the rate control unit 102 A rate S102 necessary and sufficient for the device 21 to generate a decoded image is estimated. The method for generating the encoder predicted image is not limited. For example, the encoder predicted image is generated by performing motion compensation on the reference image PIC2.

  For example, in the case of rate control as described in Non-Patent Document 2, when calculating the error statistic between the encoding target image PIC1 and the encoder predicted image, the error for each pixel is expressed as α in the translational motion region. Evaluate by double, and evaluate by β times in the non-translational motion region. Based on the error statistic obtained in this way, a rate S102 necessary and sufficient for the decoder to decode is estimated. For example, a value of 0.0 ≦ α <1.0 is used for α, and β = 1.0 is used for β.

  Further, for example, when rate control is performed as described in Non-Patent Document 3, one bit inversion generated in the translational motion region is calculated when counting the number of bit inversions between the encoding target image PIC1 and the encoder predicted image. The number of bit inversions is counted by replacing γ bit inversions and replacing one bit inversion generated in the non-translational motion region with δ bit inversions. Based on the bit inversion probability that can be calculated from the number of bit inversions obtained in this way, the rate S102 necessary and sufficient for the decoder to decode is estimated. For example, a value of 0.0 ≦ γ <1.0 is used for γ, and δ = 1.0 is used for δ.

  By applying the parameters α and β, γ, and δ as described above, in the translational motion region, the error is evaluated to be smaller than the error that actually occurs between the encoding target image PIC1 and the encoder predicted image. It will be.

  As an example of a specific calculation formula, a calculation formula for obtaining the rate S102 (“R” represents the rate in the formula (2)) based on the technique described in Non-Patent Document 3 is the formula (2). Shown in P in the equation (2) is a bit inversion probability calculated according to the equation (1). In equation (1), the total number of bits is E, the number of bit inversions occurring in the translational motion region is C1, and the number of bit inversions occurring in the non-translational motion region is C2. T is a constant determined empirically.

p = (α × C1 + β × C2) / E (1)
R = −pXlog (p) − (1−p) × log (1−p) + T (2)
The difference between the technique described in Non-Patent Document 3 and the example according to the first embodiment described above is whether or not the bit inversion numbers C1 and C2 are added with weights when the bit inversion probability p is obtained. The bit inversion probability p obtained by simply adding the bit inversion numbers C1 and C2 is a bit inversion probability reflecting an error between the encoding target image PIC1 and the encoder predicted image, and is originally intended to be estimated. This is different from the bit inversion probability reflecting the error between PIC1 and the predicted decoder image. By evaluating the bit inversion numbers C1 and C2 with weights using weighting coefficients γ and δ that take different values in the translational motion region and the non-translational motion region, the bit inversion probability p is determined as the encoding target image PIC1 and the decoder. High accuracy of rate control is realized by approaching the bit inversion probability with the predicted image.

  Even in the case of rate control as described in Non-Patent Document 2, when calculating the error statistic between the encoding target image PIC1 and the encoder predicted image, the translational motion region is evaluated by α times, and the nontranslational motion is calculated. In the region, rate control is performed by making the error statistic between the encoding target picture PIC1 and the encoder predicted picture close to the error statistic between the encoding target picture PIC1 and the decoder predicted picture so as to evaluate by β times. Realization of high accuracy.

  The translational motion region is very likely to include the same part between the encoding target image PIC1 and the reference image PIC2, and the pixels in the translational motion region have substantially the same motion vector. This is an area that is easy to predict.

  That is, it can be said that this is an area where the decoder predicted image is likely to be of high quality.

  For this reason, even if a large error occurs between the encoding target image PIC1 and the predicted encoder image, there is a possibility that the error amount between the encoding target image PIC1 and the predicted decoder image becomes small. High compared to the area. In order to reflect such a hypothesis in rate control, as described above, when calculating the error statistic between the encoding target image PIC1 and the encoder predicted image, or when counting the number of bit inversions, translation is performed. An error between the encoding target image PIC1 and the encoder predicted image is evaluated while weighting the motion region and the non-translational motion region.

  The error statistic and the bit inversion number obtained by the first embodiment can also be regarded as the error statistic and the bit inversion estimation value between the encoding target image PIC1 and the predicted decoder image.

  As described above, conventionally, the decoder has estimated a rate necessary and sufficient for decoding based on the error statistic and the bit inversion number between the encoding target image PIC1 and the encoder predicted image. In this embodiment, based on the translational motion region identification signal, an error statistic between the encoding target image PIC1 and the decoder predicted image and an estimated value of the bit inversion number are obtained, and the decoder decodes based on the estimated value. Necessary and sufficient rate S102 is estimated.

  That is, the accuracy of rate control is improved by adding a process for estimating the error statistic and the bit inversion number between the encoding target picture PIC1 and the predicted decoder picture.

  In the Wyner-Ziv encoding unit 103, the encoding target image PIC1 is Wyner-Ziv encoded according to the rate S102, and a bit stream BS is output (step ST1O8).

(A-3) Effect of the First Embodiment As described above, according to the first embodiment, the translational motion region and the non-translational motion region are identified, and the encoding target is weighted according to the difference in the region. By comparing the image with the predicted image, a rate close to a necessary and sufficient true value can be estimated on the decoder side based on the hypothesis that the quality of the decoder predicted image is high in the translational motion region. Encoding efficiency can be improved by improving the accuracy of the rate.

  Moreover, since all the processes for obtaining the translational motion region are performed by using the reduced image, even if the rate control of the first embodiment is applied, the amount of calculation is slight.

(B) Second Embodiment Next, a second embodiment of the moving picture coding apparatus and program and the moving picture coding system according to the present invention will be described with reference to the drawings.

(B-1) Configuration of Second Embodiment The second embodiment is different from the first embodiment in the integration unit in the video encoding device. That is, the moving image encoding device of the second embodiment is equivalent to the integration unit 110 in the moving image encoding device 11 of the first embodiment replaced with the integration unit 210 of the second embodiment.

  Therefore, in the following, details of the integration unit 210 of the second embodiment will be described. FIG. 5 is an explanatory diagram illustrating a detailed configuration of the integration unit 210 according to the second embodiment.

  The integration unit 210 includes a translational motion region identification signal candidate generation unit 211, an area calculation unit 212, and a translational motion region determination unit 213.

  The translational motion region identification signal candidate generation unit 211 includes a motional region identification signal S107 given from the motional region identification unit 107 (see FIG. 1) and a translational region identification unit 109 (a translational region identification signal S109 given from FIG. 1). Based on the above, a translation motion region identification signal candidate S211 is generated and output to the area calculation unit 212 and the translation motion region determination unit 213.

  The area calculation unit 212 outputs the area S212 of the candidate S211 to the translational motion region determination unit 213 based on the translational motion region identification signal candidate S211 from the translational motion region identification signal candidate generation unit 211.

  The translational motion region determination unit 213 obtains the translational motion region identification signal S101 based on the translational motion region identification signal candidate S211 from the translational motion region identification signal candidate generation unit 211 and the area S212 from the area calculation unit 212, and rate. It outputs to the control part 102 (refer FIG. 1).

(B-2) Operation of the Second Embodiment Next, the operation of the integration unit 21 in which the video encoding device of the second embodiment is different from the video encoding device 11 of the first embodiment. This will be described with reference to the drawings.

  The discriminating motion region discriminating unit in the moving picture coding apparatus according to the second embodiment also performs the operations of steps ST101 to ST105 as in the case of the first embodiment (see FIGS. 1 and 2 described above).

  That is, when a new encoding target image PIC1 becomes a processing target, the encoding target image reducing unit 104 generates a reduced encoding target image S104 from the encoding target image PIC1, and the reference image reducing unit 105 Then, a reduced reference image S105 is generated from the reference image PIC2 (step ST101), and the motion vector estimation unit 106 generates a motion vector S106 for each block from the reduced encoding target image PIC104 and the reduced reference image PIC105 (step ST1O2). ) The motion region identification unit 107 generates a motion region identification signal S107 for each block from at least one of the reduction encoding target image S104, the reduced reference image S105, and the motion vector S106 (step ST1O3), and a translation region motion vector. Calculation part In 08, a translation region motion vector S108 for identifying a translation region is generated from the motion vector S106 (step ST104). In the translation region identification unit 109, a translation region identification signal 109 is obtained from the motion vector S106 and the translation region motion vector S108. Is generated (step ST1O5).

  Thereafter, the operation of the integration unit 210 in the second embodiment is executed (step ST206). FIG. 6 is a flowchart showing a detailed operation of the integration unit 210 in the second embodiment.

  The translational motion region candidate generation unit 206 generates a translational motion region identification signal candidate S211 from the motion region identification signal S107 and the translational region identification signal S109 (step ST206-1). The method for generating the translation motion region identification signal candidate S211 is the same as the method for the integration unit 110 in the first embodiment to generate the translation motion region identification signal S101.

  In area calculation unit 212, area S212 of the region determined as the translational motion region by translational motion region identification signal candidate S211 is obtained, and the obtained area S212 is output (step ST206-2).

  In addition, although it is set as an area here, you may obtain | require the ratio of the area determined to be a translational motion area | region. By obtaining the ratio, it is possible to set the threshold value used for threshold determination in step ST206-3 described later to a value that does not depend on the frame size to be encoded.

  If the area S212 is equal to or greater than a predetermined value in the translational motion region determination unit 212, the translational motion region identification signal candidate S211 is output as the translational motion region identification signal S101, while if the area S212 is less than a certain value, A translational motion region identification signal S101 in which all regions are non-translational motion regions is output (step ST206-3).

  The operations of the rate control unit 102 and the Wyner-Ziv encoding unit 103 executed after this are the same as those in the first embodiment (see steps ST106 and ST107 in FIGS. 1 and 2 described above).

  That is, when the translational motion region identification signal S101 is generated, the bit stream BS transmitted to the video decoding device 21 from the encoding target image PIC1, the reference image PIC2, and the translational motion region identification signal S101 in the rate control unit 102. Rate (encoded data amount) S102 is estimated (step ST107), and the Wyner-Ziv encoding unit 103 performs Wyner-Ziv encoding on the encoding target image PIC1 according to the rate S102, and outputs a bitstream BS. (Step ST1O8).

(B-3) Effects of the Second Embodiment As described above, according to the second embodiment, when the area recognized as the translational motion region is smaller than the predetermined area, all Since the region is a non-translational motion region, it is possible to mitigate or prevent the adverse effect of false detection caused by noise or the like.

  As described in the operation section of the first embodiment, estimating the translational motion region using only the reduced image is one of the points (features), and an increase in the amount of calculation is suppressed. However, it is difficult to determine the motion area as a price.

  Although the motion area identification unit 107 described in the first embodiment has a function of identifying a motion area with high accuracy even under this condition, it is still an imaging apparatus even though it is not a motion area originally. In some cases, it may be determined as an exercise region due to noise or the like. Since such false detection caused by noise or the like is only one-off, the area recognized as the translational motion area in the integration unit is less than the area specified in advance as in the second embodiment. In the case where the area is also small, the influence of the above problem can be alleviated or prevented by making all the areas non-translational movement areas.

(C) Other Embodiments In the description of each of the above-described embodiments, various modified embodiments have been referred to. However, modified embodiments as exemplified below can be given.

(C-1) In each of the above embodiments, the translational motion region identifying unit has identified the translational motion region after reducing the encoding target image and the reference image, but the present invention is not limited to this. It is not something. For example, if the encoding target image and the reference image to be handled are small in size, the translational motion region may be identified without reducing the encoding target image and the reference image.

(C-2) In each of the above-described embodiments, what is identified in two stages, ie, a translational motion region or a non-translational motion region, is shown. That is, the translational motion region in which an error is likely to occur is identified between the encoder predicted image and the decoder predicted image, and the translational motion region may be identified in a plurality of stages if it can be reflected in the rate.

  For example, in the second embodiment, instead of changing a translational motion area whose area is a predetermined value or less to a non-translational motion region, it is set as a weak translational motion region, with an intermediate weight between the translational motion region and the non-translational motion region. You may make it reflect in a rate.

  Further, for example, the translational motion region obtained as a result of the determination as the motion region by the method M1 mentioned in the description of the first embodiment is set as the strong translational motion region, and the motion region is determined by any of the methods M2 to M4. The translational motion region obtained as a result of the determination may be a weak translational motion region and reflected in the rate generated by changing the weight.

(C-3) In each of the above embodiments, the motion vector estimation unit and the rate control unit are shown to be independent. However, the rate control unit uses the motion vector generated by the motion vector estimation unit to perform encoder prediction. An image may be generated. In this way, it is possible to aggregate the amount of calculation related to motion vector detection and reduce the amount of calculation.

(C-4) Although the details of the Wyner-Ziv encoding by the Wyner-Ziv encoding unit 103 have not been described in the description of each embodiment, the Wyner-Ziv encoding by the Wyner-Ziv encoding unit 103 is Any existing method may be applied as long as an encoded data amount according to a predetermined rate can be obtained.

  For example, orthogonal quantization such as Discrete Cosine Transform (DCT; Discrete Cosine Transform) may be performed after quantization, or may not be performed. In other words, it may be a moving picture coding apparatus that processes data in a pixel area (pixel-domain), or a moving picture coding apparatus that processes data in a transform coefficient area (transform-domain) after orthogonal transformation. Also good.

(C-5) In the description of each of the above embodiments, the encoding is Wyner-Ziv encoding. However, it is also possible to perform Slepian-Wolf encoding. That is, instead of the Wyner-Ziv encoding unit 103, a Slepian-Wolf encoding unit may be applied.

(C-6) In each of the above embodiments, the motion region and the translation region are obtained independently, and the translational motion region is specified (determined) by taking the logical product or the like. First, The translation region may be identified, the motion region may be identified only for the identified translation region, and the translation motion region may be specified (determined). Second, the motion region is identified and identified. Alternatively, the translation region may be identified for only the motion region, and the translation motion region may be specified (determined).

(C-7) In each of the above embodiments, the moving image encoding device based on the encoder rate control method has been described. However, for a moving image encoding device based on the decoder rate control method in which the encoder includes a rate control unit, The technical idea of the present invention can be applied and specific effects can be obtained.

  The purpose of adding the rate control unit to the moving picture coding apparatus based on the decoder rate control method is to reduce the number of feedbacks, for example. By adding the rate control unit, the rate of the Wyner-Ziv code sent for the first time is set to the rate obtained in each of the above embodiments, so that the necessary and sufficient rate can be approached and the number of feedback can be reduced. By reducing the number of feedbacks, the delay and the amount of decoder calculation are reduced.

(C-8) In each of the above embodiments, although the rate control is performed with weighting according to the translational motion region or the non-translational motion region regardless of the type of information in the coded image, Depending on the type of information in the image, rate control similar to the conventional rate control may be performed regardless of the identification result of the translational motion region or the non-translational motion region.

  For example, some conventional moving image coding apparatuses encode image data that has been quantized or orthogonally transformed by performing error correction processing on each bit plane of each digit. . In such a moving image encoding apparatus, the error correction code rate at which the error correction processing failure significantly affects the quality, such as the most significant bit plane (may include lower bit planes lower than the upper bit plane). When estimating, it is also effective to estimate the rate by comparing between the encoding target image and the encoder predicted image without weighting depending on whether the region is a translational motion region or a non-translational motion region.

  The technical idea of the present invention is that rate control is performed based on the assumption that the decoder predicted image in the translational motion area is higher in quality than the encoder predicted image, and there are rare cases where errors occur depending on the video. Therefore, when estimating the rate of an error correction code in which errors significantly affect subjective quality, the above-mentioned risk can be reduced by controlling the rate without weighting.

(C-9) The communication path from the moving picture encoding apparatus of each of the above embodiments to the opposing moving picture decoding apparatus is not limited to a narrowly defined communication path, and may be a broadly defined communication path. That is, not only real-time communication but also data encoded by the moving image encoding apparatus may be recorded on a recording medium, and data read out from the recording medium by the moving image decoding apparatus may be processed.

  DESCRIPTION OF SYMBOLS 1 ... Moving image encoding system, 11 ... Moving image encoding device (non-key frame encoding device), 12 ... Key frame encoding device, 21 ... Moving image decoding device, 22 ... Key frame decoding device, 101 ... Translation motion Area identifying unit 102 ... Rate control unit 103 103 Wyner-Ziv encoding unit 104 ... Coding target image reducing unit 105 ... Reference image reducing unit 106 106 Motion vector estimating unit 107 107 Motion region identifying unit 108 ... translational region motion vector calculation unit, 109 ... translational region identification unit, 110, 210 ... integration unit, 211 ... translational motion region identification signal candidate generation unit, 212 ... area calculation unit, 213 ... translational motion region determination unit.

Claims (17)

In a moving image encoding apparatus that compresses a moving image based on a distributed video encoding method,
A translational motion region identifying means for identifying a translational motion region in the coding target image from a coding target image that is a non-key frame and a reference image that is a key frame;
The rate that defines the code amount necessary and sufficient for the moving image decoding apparatus to generate a decoded image is different from the information of the translational motion region and the information of the non-translational motion region that is a region other than the translational motion region. Weighting is performed, and at least weighting is performed so that the information amount of the translational motion region is small, and the weighting is performed using the information on the translational motion region and the information on the non-translational motion region that are weighted. Rate control means;
A moving picture encoding apparatus comprising: an encoding unit that applies the determined rate to encode the encoding target picture.   The rate control means performs different weighting on the information on the translational motion region and the information on the non-translational motion region, and uses the information on the translational motion region and the information on the non-translational motion region to which the weighting is performed. The moving picture code according to claim 1, wherein a statistic of error is calculated between the target picture and an internally generated encoder prediction picture, and a rate is determined from the obtained statistic of error. Device. The rate control means counts the number of bit inversions between the encoding target image and the internally generated encoder predicted image, and between the encoding target image and a predicted decoder image that the video decoding device will form. The bit inversion probability existing in the is estimated, and the rate is determined from the estimated bit inversion probability,
The rate control means obtains the bit inversion number after replacing one bit inversion in the translational motion region with a value of 0 or more and less than 1 when counting the number of bit inversions. The moving picture encoding apparatus according to claim 1. The translational motion region identification means is
An encoding target image reduction unit that reduces the encoding target image and generates a reduced encoding target image;
A reference image reduction unit that reduces the reference image and generates a reduced reference image;
A motion vector estimation unit that obtains a motion vector using the reduced encoding target image and the reduced reference image;
Using only the motion vector, or using the reduced encoding target image and the reduced reference image, a motion region identification unit that identifies a motion region in the encoding target image;
A translation region motion vector calculation unit for obtaining a translation region motion vector that is a motion vector common to the translation region from the motion vector;
A translation region identifying unit for identifying a translation region from the motion vector and the translation region motion vector;
The moving image encoding apparatus according to claim 1, further comprising an integration unit that integrates the identification result of the motion region and the identification result of the translation region and determines the translation motion region.   5. The motion region identifying unit determines a block having a motion vector size larger than a predetermined threshold as a motion region, and determines other blocks as non-motion regions. The moving image encoding apparatus described.   The motion region identification unit evaluates the size of the difference between the reduced encoding target image and the reduced reference image for each block of a predetermined size using an index called a difference absolute value sum or a difference absolute value average. 5. The moving picture encoding apparatus according to claim 4, wherein a block in which the magnitude of the difference is larger than a predetermined threshold is determined as a motion area, and other blocks are determined as non-motion areas. .   The motion region identification unit obtains the absolute value of the difference between the reduced encoding target image and the reduced reference image for each pixel, and determines the number of pixels in which the absolute value of the obtained difference is greater than a predetermined threshold value. 5. The block according to claim 4, wherein each block having a predetermined size is counted, and a block in which the obtained number of pixels is larger than a predetermined threshold is determined as a motion region, and other blocks are determined as non-motion regions. The moving image encoding apparatus described.   The motion region identification unit obtains an AC component image of the reduced encoding target image and an AC component image of the reduced reference image, and determines an AC component image of the reduced encoding target image and an AC component of the reduced reference image. The absolute value of the difference from the image is obtained for each pixel, the number of pixels for which the absolute value of the obtained difference is greater than a predetermined threshold is counted for each block of a predetermined size, and the obtained number of pixels is predetermined. 5. The moving picture encoding apparatus according to claim 4, wherein a block larger than the threshold value is determined as a motion area, and other blocks are determined as non-motion areas. The motion region identification unit is
A first identification unit that determines a block having a motion vector size larger than a predetermined threshold as a motion area;
A second identification unit that determines whether or not the region is not a motion region by the first identification unit, using the reduced encoding target image and the reduced reference image; The moving picture coding apparatus according to claim 4, wherein: The translation region motion vector calculation unit calculates an average motion vector of motion vectors in a frame, and uses the obtained average motion vector as the translation region motion vector. The moving image encoding apparatus described. The translation region motion vector calculation unit obtains the most frequent motion vector among the motion vectors in a frame, and uses the obtained most frequent motion vector as the translation region motion vector. The moving image encoding device according to any one of the above.   The translation region identification unit compares the motion vector with the translation region motion vector for each block of a predetermined size, and determines that the vector is a translation region if the difference between the two vectors is less than a predetermined threshold. In other cases, the moving image coding apparatus according to claim 4, wherein the moving image coding device is determined as a non-translational region.   The integration unit determines a motion region and a region determined to be a translation region as a translational motion region, and determines all other regions as non-translational motion regions. The moving image encoding device according to any one of the above.   The integration unit determines that the area of the motion area and the area determined as the translation area is a translation area when the area is equal to or larger than a predetermined area, and determines all other areas as non-translation areas. The moving picture coding apparatus according to claim 4, wherein the moving picture coding apparatus is a moving picture coding apparatus. The rate control means divides the bit data of the same bit order into bit planes obtained for each pixel from the encoding target image consisting of a plurality of bits per pixel, and defines a code amount of each bit plane. The rate is determined based on the information of the uppermost bit plane as it is without weighting at least for the uppermost bitplane,
The encoding means encodes each bit plane related to the encoding target image by applying the determined rate of each bit plane. Video encoding device. A computer installed in a video encoding device that compresses video based on a distributed video encoding system,
A translational motion region identifying means for identifying a translational motion region in the coding target image from a coding target image that is a non-key frame and a reference image that is a key frame;
The rate that defines the amount of code necessary for the moving image decoding apparatus to generate a decoded image is different from the information of the translational motion region and the information of the non-translational motion region that is a region other than the translational motion region. Weighting is performed, and at least weighting is performed so that the information amount of the translational motion region is small, and the weighting is performed using the information on the translational motion region and the information on the non-translational motion region that are weighted. Rate control means;
A moving picture coding program that functions as coding means for coding the picture to be coded by applying the determined rate. In a moving image encoding system comprising a moving image encoding device that compresses a moving image based on a distributed video encoding method and a moving image decoding device that decodes encoded data from the moving image encoding device,
A moving picture coding system, wherein the moving picture coding apparatus according to claim 1 is applied as the moving picture coding apparatus.

Images