JP2006140683A - Image coder - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image coder, capable of obtaining a higher quality image than the conventional one with a small circuit scale and a small throughput. <P>SOLUTION: The image coder decides the existence of a moving component in the vertical direction of the apparatus by each frame, adaptively turns an input moving image in any block unit using the decision result, detects the horizontal movement from the turned image and a reference image to search for its moving vector, adaptively converts the searched horizontal moving vector to a vertical moving vector, compensates the movement relative to the turned reference image using the decision result while always using its moving vector as a horizontal moving vector, and estimates and codes by using the result obtained by movement compensating means. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a moving image capturing apparatus that detects a motion compensation vector in moving image encoding, and to an image encoding apparatus.

  In the case of encoding a moving image, it is general as a conventional technique to use interframe encoding in which encoding is performed using temporally preceding and following frames. This is a method of encoding as a motion vector using image correlation between frames. By using the motion vector, the encoding efficiency is improved and the data amount of the moving image can be reduced.

  As a search for a motion vector, a block matching method for searching for a highly correlated block in a reference frame based on a target block of a current frame is known. The shape of the search range and the search accuracy depend on the time required for encoding and the circuit scale.

  If the search accuracy is increased, more optimal block matching is performed, the prediction error is reduced, and encoding can be performed with a high compression rate. Conversely, if the search accuracy is lowered in order to prioritize the circuit scale and encoding time, the prediction error increases and the compression rate decreases. Compared with the case where the search accuracy is increased, the amount of data becomes large even with the same image quality. In other words, if the data amount is made equal, it is disadvantageous in terms of image quality. The accuracy of search and the image quality of a compressed image are generally considered to have a trade-off relationship.

  In order to cope with this problem, a motion detection method and apparatus and a camera-integrated recording apparatus disclosed in Patent Document 1 are disclosed.

In the motion detection method and apparatus and the camera-integrated recording apparatus disclosed in Patent Document 1, not only a reference frame that moves back and forth in time but also camera information (zoom, focus, motion sensor, etc.) is used to set a motion vector search range. It is characterized by deciding. Even with the same search accuracy, more accurate motion vectors can be detected than when a vector search is performed centering on the target block of the current frame.
JP-A-8-140087

  By using the disclosed example, it is possible to obtain a better block matching result than a normal search even in the same search range in area. However, although it is advantageous in that the search range does not have to be expanded, there arises a problem that the processing becomes complicated accordingly.

  In addition, since the information of many cameras must be handled in real time, there is a problem that the control itself becomes complicated.

  The present invention has been made paying attention to the above points. By using the motion vector detection means in the horizontal direction adaptively using the motion sensor information of the image encoding device, the processing load is reduced. An object of the present invention is to provide an image encoding apparatus that can reduce the size and increase the circuit size / encoding time and obtain a more efficient compressed image.

  In order to solve the above-described problem, an image encoding device according to claim 1 of the present application includes a motion detection unit that detects a motion of the image encoding device itself, and an image encoding from motion information detected by the motion detection unit. Vertical motion determination means for determining the presence or absence of a vertical motion component of the apparatus for each frame, and a block image that adaptively rotates the input moving image in units of arbitrary blocks using the result of the vertical direction motion determination means Rotation means, motion vector search means for detecting a motion in the horizontal direction from an image output from the block image rotation means and a rotation reference image described later, and searching for the motion vector, and detected by the motion vector search means Motion vector conversion means for adaptively converting a horizontal motion vector into a vertical motion vector, and a motion vector output by the motion vector conversion means And a motion compensation unit that performs motion compensation on the rotated reference image, which will be described later, a prediction encoding unit that performs prediction encoding using a result obtained by the motion compensation unit, and an output of the prediction encoding unit A local decoding means for generating a local decoded image, a local decoded image rotating means for adaptively rotating the local decoded image using a result of the vertical direction motion judging means and outputting a rotated local decoded image, and the rotated local Reference image storage means for storing the decoded image as a reference image is provided.

  The image encoding device according to claim 2 includes a motion detection unit that detects a motion of the image encoding device itself, and presence / absence of a vertical motion component of the image encoding device from the motion information detected by the motion detection unit. Vertical direction motion determination means for determining each frame, block image rotation means for adaptively rotating the input moving image in units of arbitrary blocks using the result of the vertical direction motion determination means, and the block image rotation means A motion vector search means for detecting a motion in a horizontal direction from an image output from the block rotation reference image described later and a motion vector search means, and a horizontal motion vector detected by the motion vector search means is adaptively applied. Motion vector converting means for converting the motion vector into a vertical motion vector, and a block described later using the motion vector output by the motion vector converting means Motion compensation means for performing motion compensation on the transreference image, prediction encoding means for predictive encoding using the result obtained by the motion compensation means, and local decoding by decoding the output of the prediction encoding means Using local decoding means for generating an image, reference image storage means for storing data output from the local decoding means, and an arbitrary block of the image stored in the reference image storage means, using the result of the vertical direction motion determination means And a reference block rotation image generating means for adaptively rotating and outputting as a reference block rotation image.

  The image coding apparatus according to claim 3 is characterized in that the vertical motion detection means detects the motion in the vertical direction from the ratio of the horizontal and vertical motions.

  The image coding apparatus according to claim 4 is characterized in that the vertical direction motion detecting means detects the vertical direction motion only when the vertical direction motion continues for a specific time.

  By using the motion sensor information of the image coding apparatus in the horizontal direction adaptively using the motion vector detection means in the vertical direction, the processing load is reduced, and the circuit scale / encoding time is not increased and the efficiency is increased. A good compressed image can be obtained.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

  FIG. 1 is a diagram showing a first embodiment.

  100 is input block image data input in units of arbitrary blocks, 101 is a block image rotation unit that adaptively rotates the input image 90 degrees counterclockwise in units of blocks, and 102 is referred to as the current block image of interest. A motion vector search unit for searching for a motion vector from the block image, 103 is a motion vector conversion unit that adaptively rotates the direction of the motion vector by 90 degrees counterclockwise, and 104 outputs motion compensation for the reference block image. An orthogonal transform / quantization unit that performs orthogonal transform and quantization on the current block of interest, 105 a local decoding unit that performs inverse transform on the orthogonal transformed / quantized data, and 106 adaptively demodulates the reference block image. Reference image rotation unit that rotates 90 degrees clockwise, 107 is a reference image frame memory that holds an arbitrary number of frames of reference images, and 109 is an image pickup unit Movement sensor for detecting movement of the 置自 body, 108 vertical motion determination unit determines the movement in the vertical direction from the motion detector, 110 is a coded data output from the image coding apparatus.

  The main flow of data will be described with reference to FIGS.

  The block image rotation unit 101 adaptively rotates the input block image 100, which is the current block of interest, by 90 degrees counterclockwise based on information from the vertical direction motion determination unit 108, and outputs the result to the motion vector search unit 102. . When there is no rotation instruction from the vertical direction motion determination unit 108, the block image rotation unit 101 outputs the input block image 100 to the motion vector search unit 102 without rotating. The motion vector search unit 102 includes a plurality of horizontal block images centered on the block image output from the block image rotation unit 101 and the reference block image corresponding to the same position as the current target block output from the reference image frame memory 107. By comparing sequentially, a motion vector in the horizontal direction is generated for the block having the highest correlation between the two.

  The motion vector conversion unit 103 adaptively converts a horizontal motion vector into a vertical motion vector in response to an instruction from the vertical direction motion determination unit 108, and outputs it to the orthogonal transform / quantization unit 104. When there is no rotation instruction from the vertical direction motion determination unit 108, the motion vector conversion unit 103 does not convert the horizontal direction motion vector into the vertical direction motion vector. When the motion sensor 109 does not detect the vertical motion of the image encoding device, the vertical motion determination unit 108 detects the block image rotation unit 101, the motion vector conversion unit 103, the orthogonal transform / quantization unit 104, and the reference image. The rotation instruction is not issued to the rotation unit 106, and the image coding apparatus does not perform rotation processing on the frame.

  The orthogonal transform / quantization unit 104 performs motion compensation on the output of the reference image frame memory 107 using the motion vector output from the motion vector conversion unit 103, and quantizes the input block image 100 after orthogonal transform. Here, when vertical motion is determined by the vertical motion determination unit 108, the reference block image output from the reference image frame memory 107 is 90 degrees counterclockwise according to an instruction from the vertical motion determination unit 108. Since it is rotated, the reference block image is rotated 90 degrees clockwise in the orthogonal transform / quantization unit 104 to restore the original, and then motion compensation / encoding is performed and output as encoded data 110.

  On the other hand, the motion sensor 109 detects the movement of the imaging apparatus itself in the horizontal direction and the vertical direction. The vertical motion determination unit 108 compares the magnitudes of horizontal and vertical motions, and when the vertical motion is large, the block image rotation unit 101, the motion vector conversion unit 103, the orthogonal transform / quantization unit 104, The reference image rotation unit 106 is instructed to switch the operation to a vertical motion vector search.

  A reference image generation method will be described using FIG. 7 as an example. FIG. 7 is a diagram showing the generation timing of the reference image according to the movement of the imaging apparatus itself. Frame numbers 1 to 4 are frames that have been picked up and determined to be horizontal movement by the vertical direction movement determination unit 108, frame numbers 5 to 12 are frames that are determined to be vertical movement, and frame numbers 13 to 17 are It is a frame determined to be a horizontal movement.

  Frame number 1 in FIG. 7 is the start of imaging, and the first frame is forcibly subjected to intra-frame coding, and is decoded by the local decoding unit 105 as a reference image. As indicated by frame number 5, the vertical motion determination unit 108 determines that the frame is moving in the horizontal direction in the immediately preceding frame, but it is determined that the current frame is moving in the vertical direction. And, conversely, as indicated by frame number 13, it was determined that the frame had moved in the vertical direction up to the previous frame in time, but the current frame was determined to be in the horizontal direction. In this case, the orthogonal transform / quantization unit 104 performs intra-frame coding using only the input block image 100 without using motion vector information.

  In addition, in order to use the generated intraframe encoded image as a new reference image, the orthogonally transformed / quantized data is decoded into block image data by the local decoding unit 105. In frame number 5, the block image data is rotated 90 degrees counterclockwise by the reference image rotation unit 106 in accordance with an instruction from the vertical direction motion determination unit 108, stored in the reference image frame memory 107, and data for one screen is stored. Retained. In the frame number 13, the block image data is accumulated in the reference image frame memory 107 without being subjected to rotation processing, and data for one screen is held. The reference image stored in the reference image frame memory 107 is appropriately read out from the motion vector search unit 102 in units of blocks and used for motion vector search.

  As described above, a reference image is newly generated in frame number 5 which is the first frame in which the vertical motion is detected and frame number 13 which is the first frame in which the vertical motion is no longer detected.

  As described above, it is possible to obtain better block matching by generating a new reference image each time the movement of the imaging apparatus itself changes.

  A specific example of the first embodiment will be described with reference to FIGS. 1, 3, 4, and 5.

  FIG. 3 is a diagram showing an example in which the image coding apparatus has only a horizontal movement. In this example diagram, forward prediction is described, but it is also possible to apply backward prediction or bidirectional prediction in the same manner.

  (A) is the current input image frame, (b) is the reference image, 301 is an enlarged view of the target block of the current input image frame, and 302 to 304 are centered on the position of the target block of the current input image frame in the reference image. It is an enlarged view of three blocks in the horizontal direction. In this example diagram, both (a) and (b) are divided into 9 × 6 54 blocks for simplicity of explanation, but the number of blocks is not particularly defined. In addition, the reason why the horizontal search range is set to three blocks is for simplification and is not particularly specified.

  An example in which the target block of the current image to be processed is 301 will be described. As shown in FIG. 3, when the image coding apparatus is moving in the right direction, the position of the subject is in a relationship such as the current image (a) and the reference image (b) which is a temporally previous frame. In this case, since there is no vertical motion in the image encoding device, no instruction is issued from the vertical motion determination unit 108. Accordingly, the block image rotation unit 101, the motion vector conversion unit 103, and the reference image rotation unit 106 do not perform the rotation process on the block data, and detect the motion vector in the horizontal direction as it is.

  In the example of this figure, the search range is the left and right blocks 301 and 304 centering on the block 303 of the reference image corresponding to the position of the current target block 301. Here, since the block 304 has the highest degree of correlation, the block 304 is selected by the motion vector search unit 102, and a motion vector only in the horizontal direction is detected. The detected motion vector is output to the motion vector conversion unit 103, but since there is no rotation instruction from the vertical direction motion determination unit 108, orthogonal transform / quantization is performed without performing vertical and horizontal motion vector conversion. Output to the unit 104. The orthogonal transform / quantization unit 104 performs motion compensation on the reference image block 304 using the horizontal motion vector, and encodes the current block of interest 301.

  Next, a case where there is a vertical movement will be described with reference to FIGS. FIG. 4 and FIG. 5 are diagrams showing an example in which the image coding apparatus has only a vertical movement. Similar to the above example, these example diagrams describe forward prediction, but it is also possible to apply backward prediction or bidirectional prediction in the same manner.

  (A) is a current input image frame, (c) is a reference image, (d) is a 90-degree rotated reference image, 401 is an enlarged view of a target block of the current image, and 501 is a 90-degree rotational enlargement of the target block of the current image. FIGS. 402 to 404 are three horizontal enlarged views centering on the position of the target block of the current image in the reference image, and 502 to 504 are vertical directions centering on the position of the target block of the current image in the reference image. It is a 90 degree rotation enlarged view of three blocks.

  FIG. 4 is a diagram showing a case where only a motion in the vertical direction is handled only by a motion vector search in the horizontal direction. (C) is a reference image, 401 is an enlarged view of the target block of the current input image frame, and 402 to 404 are three enlarged blocks in the horizontal direction centering on the position of the target block of the current input image frame in the reference image. .

  When the image coding apparatus is moving downward, the position of the subject is in a relationship as shown in the current image (a) and the reference image (c), which is a temporally previous frame. If the horizontal motion vector search is performed as it is, the blocks 402 to 404 centering on the reference block 403 corresponding to the same position as the current block of interest 401 have low image correlation. Therefore, none of the reference blocks 402 to 404 is selected, and intra-frame coding or one of these three reference blocks 402 to 404 having a low correlation is selected, which may lead to a decrease in compression rate. There is.

  Compared to FIG. 4, FIG. 5 shows that the motion vector search in the horizontal direction also corresponds to the vertical direction. (D) is a reference image rotated counterclockwise by 90 degrees when the reference image (c) of FIG. 4 is stored in the reference image frame memory 107, 501 is an enlarged view of a target block of the current input image frame, Reference numerals 502 to 504 are enlarged views of three blocks in the vertical direction around the position of the target block of the current input image frame in the reference image.

  As a method of rotating the reference image (c), a method of converting and storing coordinates when storing in the reference image frame memory 107, and changing the reading order by controlling the address of the reference image frame memory without changing the storage order. Another possible method is to add a frame memory dedicated to rotated images and use them by switching.

  When the vertical motion determination unit 108 detects vertical motion from the motion sensor 109, the block image rotation unit 101, motion vector conversion unit 103, orthogonal transform / quantization unit 104, and reference image rotation unit 106 are instructed to rotate 90 degrees. put out. In the frame immediately after the instruction is issued, the orthogonal transform / quantization unit 104 performs intra-frame coding using the frame at that time as a reference frame and transfers the frame to the local decoding unit 105. The locally decoded image is rotated 90 degrees counterclockwise by the reference image rotation unit 106 and stored in the reference image frame memory 107. The reference images stored in this way are shown in FIG.

  From the next frame, motion compensation is calculated based on the reference image (d) and the current block of interest is encoded. However, since the directions of the input image and the reference image do not coincide with each other, accurate motion compensation cannot be performed. Therefore, the block image rotation unit 101 rotates the current block image of 90 degrees counterclockwise in accordance with an instruction from the vertical direction motion determination unit 108 and aligns it with the reference image (d). The current block image 501 having the orientation adjusted in this way is 501.

  In this state, the motion vector search unit 102 detects a motion vector. The motion vector search unit 102 always detects only the vector in the horizontal direction. However, since the image itself is rotated by 90 degrees, the result is equivalent to the motion vector detection in the vertical direction. In the motion vector search unit 102, the block 504 having the highest degree of correlation is detected and adopted as the horizontal motion vector. The detected horizontal motion vector is converted into a vertical motion vector by the motion vector conversion unit 103 in accordance with a rotation instruction from the vertical direction motion determination unit 108. The orthogonal transform / quantization unit 104 always processes the motion vector in the horizontal direction, performs motion compensation / encoding on the reference image block 504 with the input block image 100, and in accordance with an instruction from the vertical direction motion determination unit 108, The encoded data 110 rotated 90 degrees around and returned to the original is output.

  By following the above procedure, it is possible to detect the motion in the vertical direction using only the motion detection means in the horizontal direction.

  FIG. 2 is a diagram showing a second embodiment.

  Reference numerals 100 to 105 and 108 to 110 are equivalent to the description of the drawings in the first embodiment. 201 is a reference image frame memory for holding a reference image for an arbitrary frame, 202 is a reference block image rotation unit for cutting out an arbitrary part of an image held in the reference image frame memory 201 and rotating it 90 degrees counterclockwise. is there.

  The main flow of data will be described with reference to FIG. The encoding flow from the input block image 100 to the orthogonal transform / quantization unit 104 is the same as that in the first embodiment.

  The block image data decoded by the local decoding unit 105 is accumulated in the reference image frame memory 201 and held as a frame image.

  When there is no rotation instruction from the vertical direction motion determination unit 108, the reference image frame memory 201 sequentially outputs a plurality of horizontal block images centered on a block corresponding to the same position as the current block of interest to the reference block image rotation unit 202. To do. The reference block image rotation unit 202 appropriately outputs the reference block image to the motion vector search unit 102 without rotating.

  When there is a rotation instruction from the vertical motion determination unit 108, the reference image frame memory 201 sequentially supplies a plurality of reference block images in the vertical direction around the block corresponding to the same position as the current block image to the reference block image rotation unit 202. Output. The reference block image rotation unit 202 rotates the reference block image 90 degrees counterclockwise, and appropriately outputs the rotated data to the motion vector search unit 102.

  On the other hand, the motion sensor 109 detects the movement of the imaging device itself in the horizontal direction and the vertical direction. The vertical motion determination unit 108 compares the magnitudes of horizontal and vertical motions, and when the vertical motion is large, the block image rotation unit 101, the motion vector conversion unit 103, the orthogonal transform / quantization unit 104, The reference image frame memory 201 and the reference block image rotation unit 202 are instructed to switch to a motion vector search operation in the vertical direction.

  A specific example of the second embodiment will be described with reference to FIGS.

  FIG. 6 is a diagram illustrating an example in which the imaging apparatus causes a vertical movement in the second embodiment. In this example diagram, forward prediction is described as in the first embodiment, but backward prediction or bidirectional prediction can also be applied by the same method.

  (A) is the current input image frame, (e) is the reference image, 601 is a 90 degree rotation enlarged view of the current image's target block, and 602 to 604 are centered on the position of the current image's target block 601 in the reference image. Three blocks in the vertical direction are enlarged, and reference numerals 605 to 607 are enlarged views of the three blocks in the vertical direction around the block corresponding to the same position as the current block of interest 601. As in the first embodiment, in this example, both (a) and (e) are divided into 9 × 6 54 blocks for simplicity of explanation, but the number of blocks is not particularly specified. In addition, the reason why the horizontal search range is set to three blocks is for simplification and is not particularly specified.

  As shown in FIG. 6A, when the motion of the image coding apparatus is vertically downward, the vertical motion determination unit 108 sends the block image rotation unit 101, the motion vector conversion unit 103, and the reference block image rotation unit 202 to each other. In contrast, a counterclockwise 90 degree rotation instruction is issued.

  The reference block image rotation unit 202 extracts three blocks from the reference data held in the reference image frame memory 201 as indicated by 602 to 604 in the vertical direction around the block 603 corresponding to the same position as the current block of interest 601. Are rotated by 90 degrees counterclockwise as indicated by reference numerals 605 to 607 in accordance with instructions from the vertical direction motion determination unit 108 and sequentially output to the motion vector search unit 102 as reference data.

  The block image rotation unit 101 rotates the current block of interest 90 degrees counterclockwise as indicated by 601 in accordance with an instruction from the vertical direction motion determination unit 108 to match the direction with the reference block, and the motion vector search unit 102 determines the motion vector. To detect.

  Although the motion vector search unit 102 has only a horizontal vector detection function, the image itself is rotated by 90 degrees, and as a result, vertical motion vector detection is performed. In the motion vector search unit 102, the block 607 having the highest correlation is detected and adopted as the horizontal motion vector. The detected horizontal motion vector is converted into a motion vector in the vertical direction by the motion vector conversion unit 103 and then output to the orthogonal transform / quantization unit 104.

  The orthogonal transform / quantization unit 104 performs motion compensation / encoding of the block 604 before rotation of the block 607 selected by the motion vector detection in the reference image frame memory 201 and the input block image 100, and the encoded data 110 Output as.

  By following the above procedure, it is possible to detect the motion in the vertical direction using only the motion detection means in the horizontal direction.

  The feature of the second embodiment is that the reference image frame memory 201 is brought to the previous stage of the reference block image rotation unit 202. The reference image frame memory 201 always stores the reference image before rotation, and the reference block image rotation unit 202 partially performs rotation processing as necessary. For this reason, it is not necessary to recreate a reference image every time vertical movement occurs in the imaging device, and the reference image up to that point can be used as it is and can be encoded efficiently. However, unlike the first embodiment, the reference block image rotation unit 202 always needs to perform block rotation processing while an instruction from the vertical direction motion determination unit 108 is given.

  A third embodiment will be described. In the third embodiment, even when the vertical direction motion is detected by changing the determination criterion of the vertical direction motion determination unit 108 in FIGS. 1 and 2, it is more than a certain amount compared to the amount of horizontal motion. It is configured to issue a rotation instruction for the first time.

  When an imaging device such as a video camera is used, it is considered that there is more horizontal movement than vertical movement. The movement in the vertical direction is caused by an unintentionally small camera shake that is not intended, and the number of scenes in which the vertical movement is intentionally performed is smaller than that in the horizontal movement. Most cameras have a camera shake correction function, and even if camera shake occurs, there are many cases where the recorded image is not affected.

  When such an image is input to the image coding apparatus of the present invention, the motion sensor 109 detects a motion even though the image itself does not have a motion in the vertical direction. Therefore, the vertical direction motion determination unit 108 issues a useless rotation instruction. In particular, in the first embodiment, when a vertical motion is detected, a new reference image is created. Therefore, the amount of encoded image data consumption is increased each time, and the overall compression rate can be reduced. There is sex.

  Therefore, the vertical motion determination unit 108 is provided with a function for comparing the magnitude of horizontal / vertical motion, and the detected amount of motion in the vertical direction is less than a specific level or less than that in the horizontal direction. In such a case, a mechanism that does not issue a rotation instruction is provided. In this way, more stable motion vector determination is possible.

  A fourth embodiment will be described. As described in the third embodiment, when an imaging apparatus such as a video camera is used, it is certain that there is a lot of movement in the horizontal direction, but there may be intentional movement in the vertical direction. . For example, when imaging a high building or the like nearby, or tracking a specific object such as an airplane or a bird flying in the sky.

  In the third embodiment, when the vertical motion is below a specific level or when the ratio of horizontal / vertical motion is below a certain level, the search is not switched to the vertical motion vector search. Even when it is intentionally performed, there is a possibility that the search is not switched to the motion vector search in the vertical direction. 1 and FIG. 2, the vertical motion determination unit 108 further switches to the vertical motion vector search when a vertical motion of a specified level or more is continuously generated in the image coding apparatus. This mechanism is provided.

  By adopting the above configuration, an accurate search direction can be instructed even when intentional vertical panning, camera shake, and lateral movement are mixed.

1 is a block diagram of an image coding apparatus according to a first embodiment of the present invention. The block diagram of the image coding apparatus of 2nd Example of this invention. Diagram showing horizontal cutout block when horizontal acceleration occurs Diagram showing the horizontal cutout block when vertical acceleration occurs A diagram showing a block cut out by rotating the reference image The figure which showed rotating the block cut out from the reference picture The figure showing the generation timing of the reference image according to the movement of the imaging device itself

Explanation of symbols

100 input block image 101 block image rotation unit 102 motion vector search unit 103 motion vector conversion unit 104 orthogonal transform / quantization unit 105 local decoding unit 106 reference image rotation unit 107 reference image frame memory 108 vertical direction motion determination unit 109 motion sensor 110 Encoded data

Claims (4)

In an image encoding device that encodes an input moving image obtained by imaging a subject,
Motion detection means for detecting the motion of the image encoding device itself;
Vertical direction motion determination means for determining, for each frame, the presence or absence of a vertical motion component of the image encoding device from the motion information detected by the motion detection means;
Block image rotation means for adaptively rotating the input moving image in arbitrary block units using the result of the vertical direction motion determination means;
Motion vector search means for detecting a motion in the horizontal direction from an image output from the block image rotation means and a reference image described later, and searching for the motion vector;
Motion vector conversion means for adaptively converting a horizontal motion vector detected by the motion vector search means into a vertical motion vector;
Motion compensation means for performing motion compensation on a rotated reference image to be described later using the motion vector output by the motion vector conversion means as a horizontal motion vector at all times;
Predictive encoding means for predictively encoding using the result obtained by the motion compensation means;
Local decoding means for decoding the output of the predictive encoding means to generate a local decoded image;
Local decoded image rotation means for adaptively rotating the local decoded image using the result of the vertical direction motion determination means and outputting a rotated local decoded image;
Reference image storage means for accumulating either the rotated local decoded image or an image that has not been subjected to the rotation process as a reference image;
An image encoding apparatus comprising: In an image encoding device that encodes an input moving image obtained by imaging a subject,
Motion detection means for detecting the motion of the image encoding device itself;
Vertical direction motion determination means for determining, for each frame, the presence or absence of a vertical motion component of the image encoding device from the motion information detected by the motion detection means;
Block image rotation means for adaptively rotating the input moving image in arbitrary block units using the result of the vertical direction motion determination means;
Motion vector search means for detecting horizontal motion from an image output from the block image rotation means and a reference block rotation image, which will be described later, and searching for the motion vector;
Motion vector conversion means for adaptively converting a horizontal motion vector detected by the motion vector search means into a vertical motion vector;
Motion compensation means for performing motion compensation on a block rotation reference image to be described later using the motion vector output by the motion vector conversion means as a horizontal motion vector at all times;
Predictive encoding means for predictively encoding using the result obtained by the motion compensation means;
Local decoding means for decoding the output of the predictive encoding means to generate a local decoded image;
Reference image storage means for storing data output from the local decoding means;
A reference block rotation image generation means for adaptively rotating any block of the image stored in the reference image storage means using the result of the vertical direction motion determination means and outputting it as a reference block rotation image;
An image encoding apparatus comprising: The image encoding device according to claim 1 or 2,
An image coding apparatus characterized in that the vertical motion detection means detects a vertical motion from a ratio of horizontal and vertical motions. The image encoding device according to any one of claims 1, 2, and 3,
An image coding apparatus characterized in that the vertical motion detection means detects a vertical motion only when the vertical motion continues for a specific time.

Images