JP2017103717A - Image encoding device, image decoding device, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoding device capable of obtaining satisfactory encoding efficiency.SOLUTION: The image encoding device comprises: a division part which divides an image to be encoded and generates an encoding target area; a filter part which removes noise from a decoded image area by using a nonlinear filter; and a prediction part which obtains a prediction image by searching for an area similar to the encoding target area from within an image having the noise removed therefrom by the filter part.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a video encoding device, a video decoding device, and a program.

  As a compression encoding method for moving images and still images, the image is divided into blocks, each block is subjected to a transform such as Discrete Cosine Transform (DCT), and the resulting transform coefficient is quantized and quantized. There is a method of entropy encoding the transformed coefficients. At this time, prior to conversion such as DCT, a pixel value pattern of the block is predicted by referring to a frame at a time different from the frame to which the block to be converted belongs, or pixel value information in another block of the same frame, and the prediction The residual of the pixel value sequence of the conversion target block with respect to the pixel value sequence of the predicted image as a result is obtained, and conversion such as DCT is performed on the residual value sequence.

  As a prediction method described above, there is a motion compensation prediction method that refers to a frame (picture) or slice different from the frame to which the target block belongs and uses motion information between frames. Also, in-screen prediction that refers to pixel values of blocks spatially adjacent to the target block within the frame to which the target block belongs, and predicts the pattern in the block by interpolation or extrapolation based on the directionality of the texture, etc. The system is also adopted.

Furthermore, an intra block copy method that uses a similar pattern as a predicted value of the target block in a frame to which the target block belongs, and an improved method (for example, Patent Document 1) have been proposed. Yes.
Also, a method of searching for a pattern similar to the target block in an image obtained by reducing the frame to which the target block belongs in the horizontal or / and vertical direction, and using the similar pattern as a predicted value of the target block (for example, Patent Document 2) Has also been proposed.

Japanese Patent Laying-Open No. 2015-82714 Japanese Patent No. 5321439

  However, in the conventional prediction methods such as the motion compensation prediction method, the intra prediction method, and the intra block copy method described above, the decoded image region is referred to. That is, since an image in which signal degradation has occurred particularly in a high-frequency region due to processing such as quantization is referred to, the predicted image also includes a unique waveform (for example, ringing or mosquito noise) due to the degradation. For example, the intra block copy method is particularly effective for solid-screen content (computer monitor screens, computer graphics, etc.), but if the referenced decoded image contains deterioration such as ringing, a solid-colored predicted image It becomes difficult to get. As a result, a prediction residual is generated and the encoding efficiency is lowered.

  The present invention has been made in view of such circumstances, and provides a video encoding device, a video decoding device, and a program capable of obtaining good encoding efficiency.

(1) The present invention has been made to solve the above-described problem, and one aspect of the present invention is to divide a video to be encoded and generate an encoding target region, and a decoded image. A filter unit that performs noise removal on a region by a non-linear filter, and a prediction unit that searches a region similar to the encoding target region from an image from which the filter unit has performed noise removal, and obtains a prediction image. A video encoding device provided.

(2) Further, another aspect of the present invention is the video encoding device according to (1), wherein the filter unit performs resolution conversion in addition to the noise removal, and the prediction unit In addition to noise removal, a region similar to the encoding target region is searched from the image subjected to the resolution conversion, and is used as the predicted image.

(3) According to another aspect of the present invention, a reference position acquisition unit that acquires reference information indicating a reference position for acquiring a predicted image, and noise removal using a nonlinear filter are performed on the decoded image region. The video decoding device includes a prediction unit that extracts an image at a position indicated by the reference information from the image and sets it as a predicted image.

(4) According to another aspect of the present invention, there is provided the video decoding device according to (3), wherein the prediction unit includes the reference information from an image subjected to resolution conversion in addition to the noise removal. The image at the position indicated by is extracted as the predicted image.

(5) According to another aspect of the present invention, a computer divides a video to be encoded, generates a coding target region, and performs noise removal using a nonlinear filter on the decoded image region. A program for searching for a region similar to the encoding target region from a filter unit and an image from which noise has been removed by the filter unit, and causing the region to function as a prediction unit.

(6) Further, according to another aspect of the present invention, the computer performs a noise removal using a non-linear filter on a decoded image region, a reference position acquisition unit that acquires reference information indicating a reference position for acquiring a predicted image. It is a program for extracting an image at a position indicated by the reference information from the applied image and causing it to function as a prediction unit as a predicted image.

  According to the present invention, good coding efficiency can be obtained.

It is a schematic block diagram which shows the structure of the video coding apparatus 1 by the 1st Embodiment of this invention. It is a schematic block diagram which shows the structure of the video decoding apparatus 2 by the embodiment. It is a schematic block diagram which shows the structure of the video coding apparatus 1a by the 2nd Embodiment of this invention. It is a schematic block diagram which shows the structure of the video decoding apparatus 2a by the embodiment. It is a schematic diagram of the example of an image of the encoding target by the same embodiment. It is a schematic diagram explaining the geometric transformation by the embodiment.

[First Embodiment]
Hereinafter, a first embodiment of the present invention will be described with reference to the drawings. 1 and 2 are schematic block diagrams showing configurations of a video encoding device 1 and a video decoding device 2 according to the present embodiment, respectively. The video encoding device 1 is a device that compresses and encodes video, and the video decoding device 2 is a device that decodes video that has been compression-encoded by the video encoding device 1. The video may be a still image or a moving image. The video encoding device 1 includes a block division unit 31, a subtraction unit 32, a conversion unit 33, a quantization unit 34, an entropy encoding unit 35, an inverse quantization unit 36, an inverse conversion unit 37, an addition unit 38, a buffer unit 39, A filter unit 40 and a prediction unit 41 are included.

  The block dividing unit 31 (dividing unit) spatially divides the frame (or still image) of the input video signal Pi into subsequent processing units to generate a block sequence. For example, the block dividing unit 31 divides the data into a fixed shape (for example, a rectangle having a fixed size such as 8 pixels in both vertical and horizontal directions). Alternatively, the block dividing unit 31 may be capable of being divided into partial areas having different sizes and shapes. Here, out of the block string divided by the block dividing unit 31, a block to be processed from now on is called a target block (encoding target region). The block division unit 31 sequentially inputs the blocks constituting the block string to the subtraction unit 32 and the prediction unit 41 as target blocks.

  The subtraction unit 32 subtracts each pixel value of the prediction block input from the prediction unit 41 described later from each pixel value of the target block input from the block division unit 31, and sets the result as a residual block. The subtraction unit 32 inputs this residual block to the conversion unit 33. The conversion unit 33 performs conversion processing on the residual block input from the subtraction unit 32. The conversion unit 33 inputs the conversion coefficient obtained by this conversion process to the quantization unit 34. This conversion process is, for example, discrete cosine transform, discrete sine transform, an integer approximation of these (integer transform), or Karhunen-Loeve transform.

  The quantization unit 34 quantizes the transform coefficient input from the transform unit 33. The quantization unit 34 uses the result of the quantization as a quantization index and inputs it to the entropy encoding unit 35 and the inverse quantization unit 36. The quantization step in this quantization may change for each transform coefficient. For example, when the conversion processing in the conversion unit 33 is discrete cosine transform, discrete sine transform, or integer conversion thereof, a smaller quantization step is assigned to the direct current component and the low frequency component than to the high frequency component. Also good.

  The entropy encoding unit 35 applies entropy encoding to the quantization index (quantized transform coefficient) input from the quantization unit 34 and a reference vector input from the prediction unit 41 described later. Then, data compression is performed to generate a bit string. This bit string becomes the output data Do of the video encoding device 1. The entropy coding in the entropy coding unit 35 includes, for example, Huffman coding, derivation of CAVLC (Context-Adaptive Variable Length Coding), arithmetic coding, and derivation of CABAC (Context-Adaptive Binary Arithmetic Coding). Any entropy coding scheme can be used.

  The inverse quantization unit 36 returns the transform coefficient to the transform coefficient by multiplying the quantization index input from the quantization unit 34 by a quantization step (including a quantization error corresponding to the quantization step). The inverse quantization unit 36 inputs this transform coefficient to the inverse transform unit 37. Note that the quantization step in the inverse quantization unit 36 is equal to the quantization step in the quantization unit 34. The inverse transform unit 37 restores a residual block (including an error caused by quantization) by applying inverse transform of the transform in the transform unit 33 to the transform coefficient input from the inverse quantization unit 36. . For example, when the conversion process in the conversion unit 33 is a discrete cosine transform, the inverse conversion unit 37 applies an inverse discrete cosine transform. The inverse transform unit 37 inputs the restored residual block to the adder unit 38.

  The adding unit 38 adds each pixel value of the prediction block input from the prediction unit 41 to be described later and each pixel value of the residual block input from the inverse transform unit 37, and uses the result as a decoded block. The adding unit 38 inputs this decoded block to the buffer unit 39. The buffer unit 39 is a memory that stores the decoded block input from the adding unit 38. The buffer unit 39 sequentially constructs decoded images by adding the decoded block to the target block position in conjunction with the selection of the target block in the block dividing unit 31.

  The filter unit 40 performs noise removal by a non-linear filter on the decoded image (partial image decoded until immediately before the target block position, decoded image region) recorded in the buffer unit 39, and the result is the noise-removed image. To the prediction unit 41. As noise removal by this nonlinear filter, for example, an order statistical filter (for example, a median filter) within a predetermined window can be applied. Further, as another example of noise removal by this nonlinear filter, a morphological operation (for example, an opening operation or a closing operation using an expansion operation, a contraction operation, or a combination thereof) may be applied.

  Note that the filter unit 40 can perform noise removal using a plurality of types of non-linear filters, and may switch them in image or video partial units (frame units or block units). In that case, the filter unit 40 inputs switching information indicating which type of nonlinear filter is used in each unit to the entropy encoding unit 35, and the entropy encoding unit 35 converts the switching information into a quantization index. And the reference vector. Of course, only noise removal by a single nonlinear filter may be mounted on the filter unit 40.

  The prediction unit 41 compares the target block input from the block division unit 31 with the noise-removed image input from the filter unit 40, searches for a region similar to the target block from the noise-removed image, and performs association. Do. The prediction unit 41 inputs information indicating this association to the entropy encoding unit 35 as a reference vector (reference information) indicating a reference position for acquiring a region similar to the target block. As the reference vector, for example, the image coordinates of the representative point (for example, the upper left vertex when the area is rectangular) in the noise-removed image image determined to be most similar to the target block (referred to as a corresponding area) Can be used.

  As a method for the prediction unit 41 to search for a region similar to the target block in the noise-removed image, for example, a block matching method can be used. At this time, as a method of evaluating the similarity between blocks, an index may be used in which the similarity is higher as the error between blocks (such as the sum of square errors and the sum of absolute values of errors) is smaller. You may use the parameter | index that similarity is so high that a correlation value is high. Further, as a technique for evaluating the similarity between blocks, the similarity between the feature amounts of both blocks may be evaluated. For example, a numerical value (for example, a structural similarity index (SSIM)) defined by a statistic of pixel values in a block may be used as the feature amount. Furthermore, the similarity index may be defined by linear combination or non-linear combination of a plurality of evaluation indices such as an error between blocks, a correlation, and a feature amount.

Further, the prediction unit 41 inputs a partial image (prediction image) obtained by extracting the corresponding region from the noise-removed image to the subtraction unit 32. As a result, the subtraction unit 32 becomes an image (residual block) obtained by subtracting each pixel value of the predicted image from each pixel value of the target block.
In addition, the search of the similar area in the prediction unit 41 and the determination of the switching information in the filter unit 40 are performed with the rate distortion optimum together with other units (block division unit 31, conversion unit 33, quantization unit 34, etc.) as necessary. The optimum value may be determined by performing the conversion.

  Next, the configuration of the video decoding device 2 will be described with reference to FIG. The video decoding device 2 includes an entropy decoding unit 53, an inverse quantization unit 54, an inverse conversion unit 55, an addition unit 56, a buffer unit 57, a filter unit 58, and a block transfer unit 59. The entropy decoding unit 53 (reference position acquisition unit) decodes the input data Di and restores the data compressed by the entropy code. The input data Di is output data Do output from the video encoding device 1 and input to the video decoding device 2 via a network, digital broadcast, portable recording medium, or the like. Further, the decoding in the entropy decoding unit 53 is an inverse transformation of the encoding in the entropy encoding unit 35 of the video encoding device 1, and the compression (decompression) of data compression by the entropy encoding unit 35 is performed by this decoding. A quantization index and a reference vector are obtained.

  The entropy decoding unit 53 inputs the quantization index obtained by this decoding to the inverse quantization unit 54, and inputs the reference vector obtained by this decoding to the block transfer unit 59. Since the inverse quantization unit 54 is the same as the operation of the inverse quantization unit 36, detailed description thereof is omitted. The inverse quantization unit 54 inputs the transform coefficient obtained by the inverse quantization to the inverse transform unit 55. Since the reverse conversion unit 55 is the same as the operation of the reverse conversion unit 37, detailed description thereof is omitted. The inverse transformation unit 55 inputs the residual block obtained by the inverse transformation to the addition unit 56.

  The adding unit 56 adds each pixel value of the predicted image input from the block transfer unit 59 described later and each pixel value of the residual block input from the inverse transform unit 55, and buffers the result as a decoded block. Input to the unit 57. The buffer unit 57 is the same as the operation of the buffer unit 39. When the video decoding device 2 finishes processing one frame of data from the entropy decoding unit 53, the buffer unit 57 is in a state where the decoded image of the frame is stored. The buffer unit 57 outputs this decoded image as a video image signal Po.

  The filter unit 58 performs noise removal by a non-linear filter on the decoded image (partial image decoded until just before the target block position) recorded in the buffer unit 57, and the result is sent to the block transfer unit 59 as a noise-removed image. input. The noise removal by the nonlinear filter performed by the filter unit 58 is the same as the noise removal by the nonlinear filter performed by the filter unit 40 of the video encoding device 1. Note that when the filter unit 40 of the video encoding device 1 switches and executes noise removal using a plurality of nonlinear filters, the entropy decoding unit 53 inputs the switching information obtained by decoding to the filter unit 58. The filter unit 58 applies the same filter as the nonlinear filter applied by the filter unit 40 to each unit by switching the nonlinear filter according to this switching information.

The block transfer unit 59 cuts out a partial image indicated by the reference vector information input from the entropy decoding unit 53 from the noise-removed image input from the filter unit 58. The block transfer unit 59 inputs the cut out partial image to the adding unit 56 as a predicted image.
The filter unit 58 may apply the non-linear filter only to the portion of the decoded image recorded in the buffer unit 57 that is cut out by the block transfer unit 59 as a partial image.

  As described above, the video encoding device 1 and the video decoding device 2 include the filter units 40 and 58 that apply noise removal to the decoded images, respectively. As a result, mosquito noise and ringing noise included in the predicted image can be reduced, and residuals resulting from these noises can be suppressed. Thereby, in the residual block calculated by the subtraction unit 32, the value of the residual corresponding to each pixel is suppressed, and good compression efficiency can be obtained. For example, when encoding animation, computer graphics, etc., the block of the original image corresponding to the predicted image may be substantially the same as the target block. In such a case, due to the component due to noise, It is possible to suppress the residual from becoming large.

Further, the filter units 40 and 58 include nonlinear filters, and noise removal by the filter units 40 and 58 is performed using the nonlinear filters.
Noise removal using a non-linear filter can remove noise while leaving a boundary where the color changes greatly, so it is possible to suppress residual generated by dulling the boundary and obtain good compression efficiency. .

[Second Embodiment]
The second embodiment of the present invention will be described below with reference to the drawings. In the first embodiment, when performing intra prediction or inter prediction, a noise-removed image obtained by removing noise by a non-linear filter is used for a decoded image. However, in this embodiment, noise is reduced for a decoded image. The difference is that an image subjected to resolution conversion (hereinafter referred to as a geometrically converted image) is used in addition to the removal.

  FIG. 3 is a schematic block diagram showing the configuration of the video encoding device 1a according to the present embodiment. 3, parts corresponding to those in FIG. 1 are denoted by the same reference numerals, and description thereof is omitted. The video encoding device 1a is different from the video encoding device 1 in that it has a geometric conversion unit 40a instead of the filter unit 40, and a point that switching information is input from the geometric conversion unit 40a to the entropy encoding unit 35. Is different.

  The geometric transformation unit 40a performs noise removal and geometric transformation by a non-linear filter on the decoded image (partial image decoded until just before the target block position) recorded in the buffer unit 39. The geometric conversion unit 40a inputs the image of the conversion result to the prediction unit 41 as a geometric conversion image. Noise removal by the nonlinear filter in the geometric conversion unit 40a is the same as that of the filter unit 40 in the first embodiment. Further, the geometric transformation unit 40a inputs information indicating the geometric transformation applied to the image extracted by the prediction unit 41 as a predicted image to the entropy encoding unit 35 as switching information. Note that when a plurality of types of nonlinear filters are used by switching, the geometric conversion unit 40a includes information indicating the nonlinear filter applied to each unit in the switching information. Note that information indicating only the nonlinear filter applied to the image extracted by the prediction unit 41 as the predicted image may be included in the switching information as the information indicating the nonlinear filter.

  The geometric transformation applied after the noise removal by the nonlinear filter in the geometric transformation unit 40a is, for example, reduction transformation. Prior to the reduction conversion, a spatial low-pass filter may be applied to the image. For example, this low-pass filter uses the cut-off frequency as the Nyquist frequency for resampling at the time of reduction conversion. As a filter format, for example, a Lanczos filter (Lanczos-3 filter or the like), a truncated Sinc function, a Gaussian filter, or the like can be used. The geometric conversion unit 40a may implement noise removal and geometric conversion using a plurality of types of nonlinear filters, and may switch them in image or video partial units (frame units or block units). Of course, the geometric transformation unit 40a may be mounted with only a combination of noise removal and geometric transformation by a single nonlinear filter.

  As a reference vector, the difference between the image coordinates of the representative point of the target block and the coordinates of the result of scaling and coordinate conversion of the image coordinates of the representative point of the corresponding area in consideration of the conversion in the geometric conversion unit 40a. It may be expressed by a vector. For example, when the geometric conversion unit 40a reduces the horizontal and vertical by ½, the image coordinates of the representative point of the corresponding area are doubled both horizontally and vertically. Conversely, as a reference vector, the image coordinates of the representative point of the target block between the coordinates obtained as a result of scaling and coordinate conversion in consideration of the conversion in the geometric conversion unit 40a and the image coordinates of the representative point of the corresponding region It may be expressed by a difference vector. For example, when the geometric conversion unit 40a reduces the horizontal and vertical by ½, the image coordinates of the representative point of the target area are halved both horizontally and vertically.

  FIG. 4 is a schematic block diagram showing the configuration of the video decoding device 2a according to this embodiment. 4, parts corresponding to those in FIG. 2 are denoted by the same reference numerals, and description thereof is omitted. The video decoding device 2a is different from the video decoding device 2 in that it has a geometric transformation unit 58a instead of the filter unit 58, and switching information obtained by decoding by the entropy decoding unit 53 is input to the geometric transformation unit 58a. The point is different.

  The geometric transformation unit 58a performs noise removal and geometric transformation by a non-linear filter on the decoded image (partial image decoded until just before the target block position) recorded in the buffer unit 57. The geometric conversion unit 58a inputs the image of the conversion result to the block transfer unit 59 as a geometric conversion image. Noise removal by the nonlinear filter in the geometric conversion unit 58a is the same as that in the filter unit 58 in the first embodiment. The geometric transformation in the geometric transformation unit 58a is, for example, reduction transformation. The noise removal and geometric transformation by the nonlinear filter executed by the geometric transformation unit 58a are the same as the noise removal and geometric transformation by the nonlinear filter executed by the geometric transformation unit 40a.

  When the geometric transformation unit 40a switches between noise removal and geometric transformation by a plurality of nonlinear filters, the entropy decoding unit 53 inputs the switching information obtained by decoding to the geometric transformation unit 58a. The geometric transformation unit 58a switches between the nonlinear filter and the geometric transformation in accordance with this switching information, so that the noise removal by the nonlinear filter and the geometric transformation applied by the geometric transformation unit 40a for each unit and the noise removal and the geometry by the same nonlinear transformation are applied. Apply transformation.

  An image 10 in FIG. 5 schematically represents an image to be encoded. Encoding is performed in units of blocks, and a region surrounded by a bold polygon is a region 11 that has already been encoded. The target block 12 is an area to be encoded. In general, the prediction process searches for an area similar to the pixel value pattern in the target block 12 from the encoded area 11, but the prediction process according to the present embodiment is similar to the pixel value pattern in the target block 12 ( Rotational transformation and mirror image transformation may or may not be included, and the similarity ratio may be fixed or variable).

  FIG. 5 shows an example in which a pattern having a similarity ratio twice as long as that of the target block 12 is searched from the encoded region 11 without rotational transformation and without mirror image transformation. In this example, the reference block 13 is selected as a similar region. At this time, the positional relationship between the target block 12 and the reference block 13 is also obtained. The positional relationship may be expressed, for example, as a relative vector 14 from the representative point of the reference block 13 (for example, the upper left vertex) to the representative point of the target block 12 (for example, the upper left vertex). You may express by the absolute coordinate of a representative point (for example, upper left vertex).

  The advantage of changing the resolution between the reference block and the target block as in the present embodiment will be described with reference to FIG. In this example, it is assumed that the reference block has a size that is twice as long as that of the target block. The reference block 21 is cut out from the decoded image area. For this reason, the pixel value pattern of the reference block 21 does not necessarily match the pixel value pattern of the same region 20 before encoding. In particular, ringing and mosquito noise often occur near the contour of the subject.

  On the other hand, the target block 22 associated with the reference block 21 often exhibits a step-like pixel value change with less noise near the contour, particularly in screen content such as computer graphics. Ringing and mosquito noise generated in the reference block 21 are mainly generated by quantization processing for quantizing transform coefficients such as DCT. For transform coefficients such as DCT, coefficients of high spatial frequencies tend to be roughly quantized.

  In the present embodiment, the reference block 21 is used as the prediction block 23 (predicted image) after resolution conversion. This resolution conversion mainly assumes resolution reduction (reduction conversion). When the resolution of the reference block 21 is reduced, the high frequency component included in the reference block 21 is ignored and only the component near the low frequency remains, so that the cause of ringing and mosquito noise is removed in the reference block 21. . As a result, in the prediction block 23 that is a reduced block, the vicinity of the subject outline exhibits a step-like pixel value change with less noise. As a result, the prediction residual 24 obtained by subtracting the prediction block 23 from the target block 22 becomes small.

The video encoding device 1a and the video decoding device 2a in the present embodiment can also obtain the same effects as the video encoding device 1a and the video decoding device 2 in the first embodiment.
Furthermore, the video encoding device 1a and the video decoding device 2a include geometric transformation units 40a and 58a that apply noise removal and geometric transformation to the decoded image, respectively. Thereby, the mosquito noise and the ringing noise included in the prediction block can be further reduced, and the residual due to these noises can be suppressed.

  The conversion in the geometric conversion unit 40a and the geometric conversion unit 58a may include a phase shift with an accuracy of less than one pixel in addition to the resolution conversion. As a result, more detailed alignment is possible, and the residual is reduced or smoothed by the smoothing action of the coding deterioration by the phase shift filter, so that the compression effect in entropy coding is enhanced and the coding efficiency is improved.

  In each of the embodiments described above, when the video is a moving image, the prediction unit 41 and the block transfer unit 59 may use intra prediction that searches for or cuts out a similar block from the same frame as the target block. Alternatively, inter prediction in which a similar block is searched for or cut out from a frame different from the target block may be used.

  Further, the computer can read a program for realizing the function of each unit of the video encoding device 1 in FIG. 1, the video decoding device 2 in FIG. 2, the video encoding device 1a in FIG. 3, and the video decoding device 2a in FIG. The video encoding device 1, 1a and the video decoding device 2, 2a may be realized by recording on a recording medium, reading the program recorded on the recording medium into a computer system, and executing the program. Here, the “computer system” includes an OS and hardware such as peripheral devices.

  The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Furthermore, the “computer-readable recording medium” dynamically holds a program for a short time like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case, and a program that holds a program for a certain period of time are also included. The program may be a program for realizing a part of the functions described above, and may be a program capable of realizing the functions described above in combination with a program already recorded in a computer system.

Also, the above-described functional blocks of the video encoding device 1 in FIG. 1, the video decoding device 2 in FIG. 2, the video encoding device 1a in FIG. 3, and the video decoding device 2a in FIG. Alternatively, some or all of them may be integrated into a chip. Further, the method of circuit integration is not limited to LSI, and implementation using a dedicated circuit or a general-purpose processor is also possible. Either hybrid or monolithic may be used. Some of the functions may be realized by hardware and some by software.
In addition, when a technology such as an integrated circuit that replaces an LSI appears due to progress in semiconductor technology, an integrated circuit based on the technology can be used.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes design changes and the like within a scope not departing from the gist of the present invention.

DESCRIPTION OF SYMBOLS 1, 1a ... Video encoding device 2, 2a ... Video decoding device 31 ... Block division part 32 ... Subtraction part 33 ... Conversion part 34 ... Quantization part 35 ... Entropy encoding part 36 ... Inverse quantization part 37 ... Inverse conversion part DESCRIPTION OF SYMBOLS 38 ... Addition part 39 ... Buffer part 40 ... Filter part 40a ... Geometric transformation part 41 ... Prediction part 53 ... Entropy decoding part 54 ... Inverse quantization part 55 ... Inverse transformation part 56 ... Addition part 57 ... Buffer part 58 ... Filter part 58a ... Geometric transformation part 59 ... Block transfer part

Claims (6)

A dividing unit that divides a video to be encoded and generates an encoding target region;
A filter unit that performs noise removal by a non-linear filter on the decoded image region;
A video encoding device comprising: a prediction unit that searches a region similar to the encoding target region from the image from which the filter unit has performed noise removal, and sets the region as a prediction image. In addition to the noise removal, the filter unit performs resolution conversion,
2. The video code according to claim 1, wherein the prediction unit searches for an area similar to the encoding target area from the image subjected to the resolution conversion in addition to the noise removal, and sets the prediction image as the prediction image. Device. A reference position acquisition unit that acquires reference information indicating a reference position for acquiring a predicted image;
A video decoding apparatus comprising: a prediction unit that extracts an image at a position indicated by the reference information from an image obtained by performing noise removal on a decoded image region using a non-linear filter, and uses the extracted image as a predicted image.   The video decoding device according to claim 3, wherein the prediction unit extracts an image at a position indicated by the reference information from an image subjected to resolution conversion in addition to the noise removal, and uses the extracted image as the predicted image. Computer
A dividing unit that divides a video to be encoded and generates an encoding target region;
A filter unit that performs noise removal by a non-linear filter on the decoded image region;
A program for searching a region similar to the encoding target region from an image from which the filter unit has performed noise removal, and causing the region to function as a prediction image. Computer
A reference position acquisition unit that acquires reference information indicating a reference position for acquiring a predicted image;
A program for extracting an image at a position indicated by the reference information from an image obtained by performing noise removal on a decoded image region using a nonlinear filter, and causing the image to function as a prediction image.