JP2013162467A - Stereoscopic image processing apparatus and stereoscopic image display apparatus - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve a natural adjustment without discomfort in a parallax amount range suitable for a user by detecting a parallax of a pair of images in an inputted image, thereby achieving a natural switching without discomfort while keeping a 3D image even when a scene change occurs.SOLUTION: When a parallax amount exceeds a limit value during a scene change, a value of adjustment amount is determined so that a parallax amount in a displayed image can immediately be equal to or lower than the limit value. When the parallax amount exceeds the limit value besides during a scene change, a value of adjustment amount is determined so that a parallax amount in a displayed image can gradually be equal to lower than the limit value.

Description

  The present invention relates to a stereoscopic image processing apparatus that receives a pair of image data constituting a stereoscopic image as input and generates an image in which the parallax of the image represented by the image data is adjusted. The present invention also relates to a stereoscopic image display device including a stereoscopic image processing device.

  In recent years, many stereoscopic image display techniques using binocular parallax have been proposed as image display techniques for a user to obtain a stereoscopic effect in a pseudo manner. This is a technique for making a video feel a three-dimensional image by showing an image viewed with the left eye and an image similar to that viewed with the right eye to the left and right eyes of the user, respectively.

  In addition, as a technique for showing different images to the left and right eyes of the user, the left-eye image and the right-eye image are switched alternately in time and displayed on the display. Separate the left and right eyes by using a method that separates the left and right fields of view temporally using glasses that close the field of view, and a barrier or lens that limits the display angle of the image displayed on the display in front of the display. Various schemes have been proposed, such as a scheme for displaying a left-eye image and a right-eye image.

  If the parallax is large in such a stereoscopic image display device, the pop-out amount or the pull-in amount becomes large, which can surprise the user. However, if the parallax becomes larger than a certain level and exceeds the fusion limit, the images of the right eye and the left eye are not fused, and a double image is seen and stereoscopic vision cannot be obtained. For this reason, there is a problem that a burden is placed on the eyes of the user.

  With respect to these problems, the digital broadcast receiver disclosed in Patent Document 1 moves the video display position to the left and right on the screen based on an instruction signal for adjusting the amount of parallax input from the outside. The amount of parallax is adjusted.

  In the stereoscopic video display device described in Patent Document 2, when switching from 2D video to 3D video by switching channels or the like, or when there is a difference in the amount of parallax between 3D video before and after switching, 2D display is temporarily performed. Thus, control is performed so that a 3D image is obtained after a certain period of time, thereby attempting to obtain natural switching.

  Further, in the stereoscopic video generation apparatus described in Patent Document 3, when the temporal change width of the maximum parallax amount is large, the parallax amount is adjusted to reduce the visual fatigue of the user.

JP 2000-78615 A (page 3-4, FIG. 3) JP-A-11-164328 (page 3-4, FIG. 2) Japanese Patent Laying-Open No. 2011-124941 (page 11, FIG. 5)

  In the method disclosed in Patent Document 1, the amount of parallax can be arbitrarily adjusted based on an instruction signal input from the outside. However, the method automatically adjusts according to the parallax of a pair of input stereoscopic images. In addition, the adjustment of the parallax amount is inconvenient because the user needs to perform an operation such as giving an instruction with the remote controller each time.

  In Patent Document 2, there is a problem that when a channel or the like is switched, 2D display is temporarily performed and 3D display is not performed. In addition, when scene changes frequently occur during 3D image input, 2D display is once displayed each time, and the user cannot fully enjoy stereoscopic display.

  In Patent Document 3, the time change width of the maximum parallax amount is monitored, and when the change width per time is large, the parallax amount adjustment is performed. However, in the case of a still image, since there is no change per time, it cannot cope with a 3D still image with an excessive amount of parallax. In addition, when the amount of parallax gradually increases and becomes an excessive amount of parallax, the amount of change in the amount of parallax per time is small, so the amount of parallax is not adjusted and cannot be handled. In addition, to reduce the load on the system, it is also described that the change width per time is monitored only at the time of a scene change, but as in the case described above, the amount of parallax of a stereoscopic image increases after a scene change. If it becomes too large, it cannot respond.

  The present invention has been made to solve the above-described problems. It detects a parallax between a pair of input images, and can naturally adjust the parallax amount range suitable for the user without a sense of incongruity. It is an object of the present invention to be able to switch without a sense of incongruity as a 3D image even when a problem occurs.

The stereoscopic image processing apparatus of the present invention is
A left-eye image represented by the left-eye input image data, the input image data for the left eye and the input image data for the right eye paired with each other. A parallax detection unit that detects a maximum value of the parallax amount between the images for the right eye represented by the input image data for the right eye or an average value over a plurality of frames of the maximum value;
A scene change detection unit for detecting a scene change in the left eye image and the left eye image;
A parallax adjustment unit that generates an adjustment amount for causing the parallax amount in the displayed image to be equal to or less than the limit value when the detected maximum value of the parallax amount exceeds a set limit value;
From the image data obtained by delaying the input image data for the left eye and the input image data for the right eye, or the input image data for the left eye and the input image data for the right eye for one or more frame periods, An image generation unit that generates display image data for the left eye and display image data for the right eye in which the parallax amount is adjusted according to the adjustment amount;
A display output unit that outputs a display image signal of a predetermined format from the display image data for the left eye and the display image data for the right eye,
The parallax adjustment unit includes:
When a scene change is detected by the scene change detection unit and the parallax amount detected by the parallax detection unit exceeds the limit value, the parallax amount in the displayed image is immediately equal to or less than the limit value. The value of the adjustment amount is determined so that
When no scene change is detected by the scene change detection unit and the amount of parallax detected by the parallax detection unit exceeds the limit value, the amount of parallax in the displayed image gradually increases to the limit value. The value of the adjustment amount is determined so as to be as follows.

  According to the present invention, the parallax between a pair of input images can be detected, and the parallax amount range suitable for the user can be adjusted naturally without any sense of incongruity, and even when a scene change occurs, the 3D image remains natural. Switching without a sense of incongruity is possible.

1 is a block diagram showing a stereoscopic image processing apparatus according to Embodiment 1 of the present invention. (A) And (b) is a figure which shows the relationship between the parallax of the input image for left eyes, and the input image for right eyes, and the feeling of depth which a user feels. It is a block diagram which shows the structural example of a parallax adjustment part. It is a table | surface which shows the part for adjustment amount change with respect to adjustment amount difference. It is a table | surface which shows the specific example for the adjustment amount change with respect to an adjustment amount difference. It is a graph which shows the specific example for the adjustment amount change with respect to an adjustment amount difference. It is a block diagram which shows the other structural example of a parallax adjustment part. It is a block diagram which shows the further another structural example of a parallax adjustment part. It is a block diagram which shows the further another structural example of a parallax adjustment part. It is a block diagram which shows the further another structural example of a parallax adjustment part. (A) And (b) is a figure which shows the movement of the image by parallax adjustment. (A) And (b) is a figure explaining the position of the object shown to the pixel and user which are shown on a display surface. (A) And (b) is a figure which shows the expansion of the image by scaling of the image after parallax adjustment. (A)-(d) is a figure which shows the example from which the method for filling the edge part on either side of a screen produced with parallax with an image differs. It is a block diagram which shows the stereo image processing apparatus which concerns on Embodiment 2 of this invention.

Embodiment 1 FIG.
FIG. 1 is a diagram showing a configuration of a stereoscopic image processing apparatus according to Embodiment 1 of the present invention. In FIG. 1, the stereoscopic image processing apparatus 100 includes a parallax detection unit 1, a scene change detection unit 2, a parallax adjustment unit 3, an image generation unit 4, and a display output unit 5, and is connected to a display unit 200. A stereoscopic image display apparatus is configured.

  In the stereoscopic image processing apparatus 100, image data of consecutive frames (frame sequence) constituting a moving image is sequentially input. The latest frame in the input frame sequence is referred to as current frame data or time n frame data, and may be simply referred to as frame n data.

First, an outline of the components will be described.
The parallax detection unit 1 is a left-eye image represented by left-eye input image data Da1 for each screen (frame) from the left-eye input image data Da1 and the right-eye input image data Db1 of a stereoscopic image. And the pop-up side maximum value Tmax (n) and the pull-in side maximum value Hmax (n) of the parallax with the right-eye image represented by the right-eye input image data parallax amount Db1 are output.

  The scene change detection unit 2 detects a scene change from the left-eye input image data Da1 and the right-eye input image data Db1, and determines the value of the scene change detection signal SC according to the detection result.

  The parallax adjustment unit 3 is set based on the parallax of the image represented by the input image data Da1 and Db1 so that the parallax of the displayed image is immediately or gradually within a parallax amount range suitable for the user. An adjustment amount (parallax amount adjustment amount) F (n) to be a parallax offset amount to be determined is determined, and is determined by the user with the pop-out side maximum value Tmax (n) and the pull-in side maximum value Hmax (n). The adjustment amount F (n) is determined and output from the pop-out side limit value ST and the pull-in side limit value SH and the scene change detection signal SC.

The image generation unit 4 receives the left-eye input image data Da1 and the right-eye input image data Db1,
Based on the adjustment amount F (n), left-eye display image data Da3 and right-eye display image data Db3 are generated.
The display output unit 5 outputs an image signal corresponding to the input format of the display unit 200 based on the left-eye display image data Da3 and the right-eye display image data Db3.
The display unit 200 displays a stereoscopic image based on the image signal output from the display output unit.

Next, the pop-out side maximum value Tmax and the pull-in side maximum value Hmax obtained by the parallax detection unit 1 will be described.
2A and 2B show the relationship between the parallax between the image for the left eye and the image for the right eye and the sense of depth that the user feels. The image for the left eye is only for the left eye of the user. It is assumed that the image for use is controlled so as to be visible only to the user's right eye.

  In FIG. 2A, the display position of the left-eye image portion corresponding to an object on the display surface is indicated by P1L, and the display position of the right-eye image portion on the display surface is indicated by P1R. Similarly, in FIG. 2B, the display position of the left-eye image portion corresponding to an object on the display surface is indicated by P2L, and the display position of the right-eye image portion on the display surface is indicated by P2R. When the user sees the image part at the position P1L and the image part at the position P1R as one object, the object is perceived as being present at the spatial position F1, and the image part at the position P2L and the image part at the position P2R are When it appears as one object, it is perceived that the object exists at the spatial position F2.

  Here, when the horizontal coordinate of the image portion for the left eye corresponding to a certain object is X (PL) and the horizontal coordinate of the image portion for the right eye is X (PR), the parallax d is obtained by the following equation (1). It is done. Here, since the number of pixels on one line in one frame is determined by the format of the input image, the horizontal coordinate is described as being expressed in units of the number of pixels with the zero coordinate as the left end of the frame. Hereinafter, not only the parallax d but also the adjustment amount F (n), the adjustment amount target value M (n), the adjustment amount difference S, and the adjustment amount change amount K described later are also expressed by the number of pixels. .

d = X (PL) −X (PR) (1)
It is obtained when equation (1) is applied to P1L and P1R in FIG. 2A. D1 = X (P1L) −X (P1R)
Becomes a positive value and is obtained when applied to P2L and P2R in FIG. 2B. D2 = X (P2L) −X (P2R)
Is negative.

  As shown in FIG. 2A, when the parallax is obtained with a positive value, since the object appears to exist between the display surface and the user, the amount of parallax when the parallax d is a positive value is calculated. It is called quantity. Further, as shown in FIG. 2B, when the parallax is obtained with a negative value, the object appears to be present behind the display surface, so that the parallax amount when the parallax d is a negative value is drawn in. That's it. Here, in an input image of one screen (hereinafter referred to as one frame), a parallax having a positive maximum value is referred to as a pop-up side maximum value Tmax, which is represented by a symbol Tmax and has a negative maximum value. This is called the maximum value and is represented by the symbol Hmax.

The parallax detection unit 1 calculates the above-described pop-out-side maximum value Tmax and pull-in-side maximum value Hmax for each frame, compares the left-eye input image with the right-eye input image, and inputs the left-eye input. A pixel displaying the same object is detected in the image and the input image for the right eye, and a parallax d between the pixels displaying each object is calculated by calculating a difference between the horizontal coordinates of the pixels. . For each frame, the maximum value when the parallax d is positive, that is, the pop-up side maximum value Tmax (n), and the maximum absolute value when the parallax d is negative, that is, the pull-in side maximum value Hmax (n) are calculated. To do.
Here, Tmax (n) is the maximum value on the pop-out side of frame n, and Hmax (n) is the maximum value on the pull-in side of frame n.

  When obtaining the parallax d, the screen is divided into certain sections, not for each pixel, and the parallax d is obtained for each section, and the pop-out side maximum value Tmax (n) and the pull-in side maximum value Hmax (n) in one screen are obtained. You may ask. As a method for detecting pixels displaying the same object, for example, a technique such as block matching can be used.

  The scene change detection unit 2 determines a scene change based on the left-eye input image data Da1 and the right-eye input image data Db1. When a scene change occurs, the scene change detection signal SC is set to a significant level, for example, “ 1 ″ and output to the parallax adjustment unit 3. When no scene change has occurred, SC = 0 is output. When a scene change occurs, the scene change detection signal SC is switched to “1” during the blanking period between the end of the video frame where the scene change occurred and the start of the video of the next frame, and the end of the next video frame. It is desirable to return the scene change detection signal SC to “0”.

  The scene change detection method in the scene change detection unit 2 is such that one frame image (an image represented by an image signal of the video valid period) is P blocks in the horizontal direction and Q blocks Bpq (p = 1 to P in the vertical direction). , Q = 1 to Q), the average luminance Ypq (n) of each divided block Bpq is calculated, compared with the average luminance Ypq (n-1) of each block one frame before, and the average A scene in which the number V of blocks whose luminance Ypq (n) −Ypq (n−1) difference is equal to or greater than a predetermined value Ypqth is equal to or greater than a threshold value Vth, for example, greater than or equal to half of the total number of blocks in a video effective period of one frame. It is determined that there is a change, and a scene change detection signal SC = 1 is output.

Further, as a simpler scene change detection method, either of the left-eye input image data Da1 and the right-eye input image data Db1 and the overall average brightness Yw (n) of the previous frame and the overall average brightness Yw (n−1) one frame before. If the difference is equal to or greater than the predetermined value Ywth, it is determined that the scene has changed, and the scene change detection signal SC = 1 is output. In other words, if there is a large average luminance change, a scene change is acceptable. Various methods for scene change detection have been devised and may be performed by known methods.
In addition, it is not determined whether only one of the left-eye input image data Da1 and the right-eye input image data Db1 has a change in the overall average brightness greater than a predetermined value, but the overall average brightness for both. If there is a change greater than a predetermined value on at least one side, it may be determined that the scene has changed, and only if there is a change greater than the predetermined value on both sides It may be judged as a scene change.

  In the above description, the scene change is obtained from the average luminance, but it may be determined that the scene change has occurred due to a large change in the amount of parallax. In this case, for example, the parallax detection unit 1 uses the pop-up side maximum value Tmax (n), the pull-in side maximum value Hmax (n), the pop-out side maximum value Tmax (n−1) one frame before, and the pull-in side. The maximum values Hmax (n−1) are respectively compared, and if there is a change equal to or greater than the threshold values Tth and Hth in at least one, it is determined that a scene change has occurred, and the scene change detection signal SC = 1. That is,

When Tmax (n) −Tmax (n−1) ≧ Tth and / or Hmax (n) −Hmax (n−1) ≧ Hth SC = 1 (2A)
Otherwise
SC = 0 (2B)

  As described above, in the case where the above-described parallax detection unit 1 and the scene change detection unit 2 are combined, the scene change detection unit 2 may be omitted. By doing so, there is an effect that the circuit can be simplified.

Next, the parallax adjustment unit 3 will be described.
For example, the parallax adjustment unit 3 is determined by the user and is input with the pop-out side limit value ST and the pull-in side limit value SH input from the outside of the apparatus, and the pop-out side maximum value Tmax (n) output from the parallax detection unit 1. The adjustment amount F (n) is determined and output from the side maximum value Hmax (n) and the scene change detection signal SC output from the scene change detector 2. Here, F (n) represents the parallax adjustment amount of frame n.

  In general, when the amount of parallax on the pop-out side or the amount of parallax on the pull-in side is large, the eyes may become tired or the stereoscopic vision may be difficult. The optimum value for the pop-out side maximum value and the pull-in side maximum value differs depending on the screen size, the number of pixels, the viewing distance of the user, and the like.

  Generally, a parallax angle within 1 degree is considered to be comfortable. The number of pixels varies depending on the viewing distance and the screen size when the parallax angle is 1 degree. For example, when the resolution is 1920 × 1080 pixels on a high-definition television (screen size ratio 16: 9), viewing is performed at a distance three times the screen height. At this time, the parallax angle of 1 degree is 57 pixels regardless of the screen size.

  On the other hand, regarding the pull-in side, it is appropriate that the amount of parallax does not exceed 5 cm of the interocular distance when the pull-in amount on the display screen is the maximum value (infinity). Since the distance between both eyes is about 5 cm for a child, it is desirable not to exceed 5 cm for a household device. For example, in the case of a 50 inch 1920 × 1080 pixel resolution, the number of pixels corresponding to 5 cm is 87 pixels. The number of pixels corresponding to 5 cm is determined by the screen size regardless of the viewing distance.

  The pop-out limit value ST and the pull-in limit value SH are determined from the screen size of the display unit 200, the number of pixels, the viewing distance of the user, the test image viewing result from the display unit 200, or past viewing experience. Yes, the limit values ST and SH may be numerically input by the user, or the user may select from a plurality of candidates.

A first method for obtaining the adjustment amount F (n) will be described. When the first method is used, the parallax adjustment unit 3 is configured as shown in the functional block diagram of FIG.
The illustrated parallax adjustment unit 3 includes an adjustment amount target value generation unit 301, an adjustment amount difference generation unit 302, an adjustment amount change generation unit 303, an adjustment amount generation unit 304, a memory 305, and switching units 306 and 307. And have.

  The switching units 306 and 307 are controlled by the scene change detection signal SC. When the value of the scene change detection signal SC is significant (when SC = 1), the switching contact C is connected to the A side, When the value of the scene change detection signal SC is not significant (when SC = 0), it is connected to the B side.

  First, the adjustment amount target value generation unit 301 determines the adjustment amount target value M (n) in the following four cases. Here, M (n) represents the adjustment amount target value of frame n.

First, as shown in the following formula (3), when the pop-out side maximum value Tmax (n) is larger than the pop-out side limit value ST and the pull-in side maximum value Hmax (n) is equal to or less than the pull-in side limit value SH, A value obtained by adding a negative sign to the value obtained by subtracting the pop-out side limit value ST from the pop-out side maximum value Tmax is defined as an adjustment amount target value M (n).
This is because when the pop-up-side parallax amount is large, the object represented by the image is pulled in (so that the parallax changes in the negative direction).

Next, as shown in the following equation (4), when the pop-out side maximum value Tmax (n) is larger than the pop-out side limit value ST and the pull-in side maximum value Hmax (n) is larger than the pull-in side limit value SH, The sum of the value obtained by subtracting the pop-out side limit value ST from the pop-out side maximum value Tmax (n) and the value obtained by subtracting the pull-in side limit value SH from the pull-in side maximum value Hmax (n). The adjustment amount target value M (n).
This is because the operation is performed so as to reduce the amount of parallax with the larger difference between the pop-out limit value ST and the pull-in limit value SH.

Next, as shown in the following equation (5), when the pop-out side maximum value Tmax (n) is not more than the pop-out side limit value ST and the pull-in side maximum value Hmax (n) is larger than the pull-in side limit value SH, A value obtained by subtracting the pull-in side limit value SH from the pull-in side maximum value Hmax (n) is set as an adjustment amount target value M (n).
This is because when the pop-out side maximum value Tmax is small (below the limit value ST) and the pull-in side maximum value Hmax is large, the parallax amount is changed from the pull-in side to the pop-out side.

  Finally, as shown in the following formula (6), when the pop-out side maximum value Tmax (n) is not more than the pop-up side limit value ST and the pull-in side maximum value Hmax (n) is not more than the pull-in side limit value SH, Let M (n) = 0. That is, when both the pop-out side maximum value and the pull-in side maximum value are within the range from the pop-out side limit value ST to the pull-in side limit value SH (limit range, that is, a parallax amount range suitable for the user), the target value is reached. And M (n) = 0.

  In the above example, when the pop-up side maximum value Tmax (n) and the pull-in side maximum value Hmax (n) are larger than the respective limit values ST and SH, as shown in Expression (4), the adjustment amount target Although the value M (n) has been determined, it is generally said that it is desirable that the maximum value on the drawing side does not exceed the interocular distance. For this reason, the adjustment of the pull-in side maximum value is given priority over the adjustment of the pop-out side maximum value, and the following equation (4B) may be changed instead of the equation (4).

Hereinafter, the case where the scene change has not occurred and the case where it has occurred will be described in order.
First, a case where no scene change has occurred will be described.
When no scene change has occurred, the scene change detection signal SC = 0. At this time, the switches 306 and 307 have their switching contacts C connected to the B side.

  When the scene change has not occurred, if the amount of parallax changes with time, a sense of incongruity is generated and a burden is placed on the eyes. Therefore, in order to prevent an abrupt change in time, a process for delaying the change in the amount of parallax is performed.

First, the adjustment amount difference generation unit 302 calculates the difference between the adjustment amount target value M (n) of the current frame (frame n) and the adjustment amount F (n−1) of the previous frame as the adjustment amount difference S. This process is represented by the following formula (7).
S = M (n) −F (n−1) (7)

Next, the adjustment amount change generation unit 303 converts the adjustment amount difference S into an adjustment amount change K. This conversion is performed using, for example, a conversion table indicating the relationship between the adjustment amount difference S and the adjustment amount change K.
As will be described later, the adjustment amount change amount K is a change width (F (n) -F (n-1)) of the adjustment amount between successive frames, that is, a width that changes by adjustment of one frame. .

An example of the conversion table constituting the adjustment amount change generation unit 303 is shown in FIG.
In FIG. 4, a1 to ak and -b1 to -bl each represent a value of S, and e1 to ei and -f1 to -fj each represent a value of K. e1 to ei and -f1 to -fj may be continuous values or discontinuous values.
In the illustrated example, one K value is defined for one or more S values. For example, when the S value is in the range from + a1 to + a2, the value is + e1. Indicates K is output. In addition, when the value of S is + ak or more, the value of K is + ei, and when the value of S is -bl or less, the value of K is -fj. As a result, the value of K is kept within the range from + ei to -bl.

FIG. 5 shows a specific example of the conversion table of FIG. FIG. 6 is a graph showing the relationship between S and K shown in FIG.
For example, if S = + 7, it is limited to K = + 2, and if S = -7, it is limited to K = -4.
The conversion characteristics shown in the figure suppress the change in K with respect to the change in the value of S, and in the pop-out direction (plus direction), in the pop-out direction (plus direction), the degree of suppression becomes stronger, so The change in the direction is made slower and the change in the pull-in direction is made faster.
In general, it is said that fusion is easy to hold if the pop-out direction is slow, and fusion can be held even if the pull-in is less than the above limit value even if the change is fast. Therefore, it is effective to control the pop-out direction slowly and the pull-in direction fast.

  In this way, when no scene change occurs, the adjustment amount change amount K is determined to be within a predetermined range based on the conversion table of FIG. 4 or FIG.

  The adjustment amount generation unit 304 includes an adjustment amount F (n−1) one frame before and an adjustment amount change K generated by the adjustment amount change generation unit 303, as shown in the following equation (8). Is added to obtain the adjustment amount F (n).

When SC = 0 F (n) = F (n−1) + K (8)
The adjustment amount F (n) obtained as described above is an output of the parallax adjustment unit 3.

  In this way, the adjustment amount F (n) is determined so as to be limited to be within a certain range when the scene change has not occurred. By controlling in this way, a sudden change in the amount of parallax is suppressed, and the amount of parallax is adjusted slowly.

  Next, a case where a scene change occurs will be described. When a scene change occurs, the scene change detection signal SC = 1 over one frame immediately after that. At this time, the switches 306 and 307 are both connected to the switching contact C on the A side, and the processes by the adjustment amount difference generation unit 302, the adjustment amount change generation unit 303, and the adjustment amount generation unit 304 are not performed. Instead, as indicated by the following equation (9), the adjustment amount target value M (n) is immediately output as it is as the adjustment amount F (n). Thereby, the parallax amount is immediately suppressed to the pop-out side limit value ST or less or the pull-in side limit value SH or less.

That is,
When SC = 1 F (n) = M (n) (9)

  In this way, when a scene change occurs, a process for switching the adjustment amount F (n) so that it immediately matches the adjustment amount target value M (n) is performed. At the time of a scene change, the contents of the displayed image are switched, so there is no sense of incongruity even if the amount of parallax changes suddenly. Conversely, if the amount of parallax is changed slowly, it will appear unnatural. Therefore, when a scene change occurs, it is desirable to reset the parallax in accordance with the screen change and immediately switch the adjustment amount F (n) to a value equal to the adjustment amount target value M (n).

Next, a second method for obtaining the adjustment amount F (n) of the parallax adjustment unit 3 will be described. When the second method is used, the parallax adjusting unit 3 is configured as shown in the functional block diagram of FIG.
The illustrated parallax adjustment unit 3 includes a pop-up side maximum value averaging unit 311, a pull-in side maximum value averaging unit 312, time width determination units 313 and 314, an adjustment amount target value generation unit 301 b, and an adjustment amount difference generation Unit 302, adjustment amount change generation unit 303, adjustment amount generation unit 304, memory 305, and switching units 306 and 307.
The same reference numerals as those in FIG. 3 denote the same members.

The time width determining unit 313 determines a time width (number of frames) Lt when the pop-up side maximum value averaging unit 311 obtains a time average. The time width determination unit 314 determines a time width (number of frames) Lh when the pull-in side maximum value averaging unit 312 obtains a time average.
The time width determination units 313 and 314 normally output the normal time widths Ltd and Lhd input from the outside as the time widths Lt and Lh, respectively. Ltd and Lhd are each an integer of 2 or more, and may be the same value or different from each other. When a scene change occurs, the time widths Lt and Lh are set to “1” only for the immediately following one frame period, and then the normal time widths Ltd and Lhd are increased by one for each frame period. After that, the value is maintained thereafter.

  The pop-out side maximum value averaging unit 311 averages the pop-out side maximum value Tmax over the past Lt frames, and outputs an average pop-out side maximum value ATmax. This averaging is represented by the following formula (10).

  Similarly, the pull-in side maximum value averaging unit 312 averages the pull-in side maximum value Hmax over the past Lh frames, and outputs an average pull-in side maximum value AHmax. This averaging is represented by the following formula (11).

  As the time widths Lt and Lh are larger, the adjustment amount F (n) changes more slowly, and accordingly, the parallax amount of the displayed image changes more slowly. On the other hand, the smaller the time widths Lt and Lh, the faster the change in the adjustment amount F (n), and the faster the change in the parallax amount of the displayed image.

  After a scene change has not occurred and a sufficient time has elapsed since the detection of the previous scene change (that is, after the Ltd and Lhd frame periods have elapsed from the previous scene change, respectively), Lt = Ltd, Lh = Lhd The pop-out side maximum value averaging unit 311 obtains the average of the pop-out side maximum value Tmax (n) over the past Ltd frames, and the pull-in side maximum value averaging unit 312 calculates the pull-in side maximum value Hmax (n). An average HTmax over the past Lhd frames is determined.

  In this way, by taking the time average of the pop-out side maximum value Tmax and the pull-in side maximum value Hmax, a sudden change in parallax and false detection can be mitigated.

When a scene change occurs, the time widths Lt and Lh are set to “1”. As a result, the outputs of the pop-out side maximum value averaging unit 311 and the pull-in side maximum value averaging unit 312 have the same values as their inputs. That is, the time average over a plurality of frames is not performed, and the maximum values Tmax (n) and Hmax (n) are set as the average pop-out side maximum value ATmax and the average pull-in side maximum value AHmax, respectively. Output from the maximum value averaging unit 312.
Thereby, the adjustment amount F (n) can be set to a value so that the parallax amount in the displayed image immediately becomes a value within the limit range (a range suitable for the user).

After the scene change, the time widths Lt and Lh are increased by 1 for each frame.
As a result, gradually return to the normal state (a state in which a sudden change in the amount of parallax for adjustment is suppressed) without being affected by the parallax before the scene change (maximum value on the pop-out side, maximum value on the pull-in side). Can do.

  Next, in the adjustment amount target value generation unit 301b, based on the average pop-up side maximum value ATmax, the average pull-in side maximum value AHmax, the pop-out side limit value ST, and the pull-in side limit value SH, the following four cases are classified: An adjustment amount target value M (n) is determined. Here, M (n) is the adjustment amount target value of frame n.

  In Expression (13), the adjustment of the pull-in side maximum value may be given priority over the adjustment of the pop-out side maximum value, and Expression (13) may be changed to the following Expression (13B).

  The operations of the adjustment amount difference generation unit 302, the adjustment amount change generation unit 303, the adjustment amount generation unit 304, and the memory 305 are the same as those in the first method. That is, when no scene change has occurred, the adjustment amount difference generation unit 302, based on the adjustment amount target value M (n) obtained as described above, performs the same as the first method described above. The difference between the adjustment amount target value M (n) and the adjustment amount F (n−1) one frame before is calculated as the adjustment amount difference S. This process is expressed by the above equation (7).

  The adjustment amount change generation unit 303 converts the adjustment amount difference S into the adjustment amount change K in the same manner as in the first method.

As shown in the above equation (8), the adjustment amount generation unit 304 includes the adjustment amount F (n−1) one frame before and the adjustment amount change K generated by the adjustment amount change generation unit 303. Is added to obtain the adjustment amount F (n).
The adjustment amount F (n) obtained as described above is an output of the parallax adjustment unit 3.

  In this way, the adjustment amount F (n) is limited to be within a certain range when the scene change has not occurred. By controlling in this way, a sudden change in the amount of parallax is suppressed, and the amount of parallax is adjusted slowly.

  When a scene change occurs, the scene change detection signal SC = 1 over one frame immediately after that. At this time, in the switching units 306 and 307, the switching contact C is connected to the A side, and the adjustment amount difference generation unit 302, the adjustment amount change generation unit 303, and the adjustment amount generation unit 304 are not processed. Instead, as shown by the above equation (9), the adjustment amount target value M (n) is immediately output as it is as the adjustment amount F (n).

  Thus, when a scene change occurs, the pop-up side maximum value Tmax (n) and the pull-in side maximum value Tmax (n) and the pull-in side maximum value Hmax (n) are not averaged over a plurality of frames. The maximum value Hmax (n) is output as it is as the average pop-out side maximum value AT and the average pull-in side maximum value AH, and the adjustment amount F (n) is immediately changed to a value equal to the adjustment amount target value M (n). .

  In this way, when a scene change occurs, the restriction on the time average process and parallax change over a plurality of frames is interrupted so that the adjustment amount F (n) immediately matches the adjustment amount target value M (n). By doing so, the amount of parallax can be immediately set to a value within the limit range without feeling uncomfortable.

  When the scene change does not occur, the pop-up side maximum value ATmax (n) and the pull-in side maximum value HTmax (n) are time-averaged over the time widths Ltd and Lhd. The adjustment amount target value M (n) may be output as the adjustment amount F (n) without processing the expressions (7) and (8). Further, when a scene change occurs, the time widths Lt and Lh in the averaging by the above formulas (10) and (11) are once set to “1” and then gradually increased, so that the adjustment amount target value M (N) may be immediately output as the adjustment amount F (n) as it is. In this case, the adjustment amount difference generation unit 302, adjustment amount change generation unit 303, adjustment amount generation unit 304, memory 305, and switching units 306 and 307 in the configuration of FIG. 7 may be omitted, and the configuration of FIG.

The processing in this case is represented by the following formula (16). That is,
In both cases of SC = 0 and SC = 1,
F (n) = M (n) (16)

Furthermore, a third method for obtaining the adjustment amount F (n) of the parallax adjustment unit 3 will be described. When the third method is used, the parallax adjusting unit 3 is configured as shown in the functional block diagram of FIG.
The illustrated parallax adjustment unit 3 includes an adjustment amount target value generation unit 301, an adjustment amount target value averaging unit 321, a time width determination unit 323, an adjustment amount difference generation unit 302b, and an adjustment amount change generation unit 303. , An adjustment amount generation unit 304, a memory 305, and switching units 326 and 307.
The same reference numerals as those in FIG. 3 denote the same members.

  Similarly to the switching unit 307, the switching unit 326 is controlled by the scene change detection signal SC. When the value of the scene change detection signal SC is significant (when SC = 1), the switching contact C is set to A. When the value of the scene change detection signal SC is not significant (when SC = 0), the switching contact C is connected to the B side.

The time width determination unit 323 determines a time width (number of frames) Lm when the adjustment amount target value averaging unit 321 obtains a time average. The time width determination unit 323 normally outputs the normal time width Lmd input from the outside as the time width Lm as it is. Lmd is an integer of 2 or more.
When a scene change occurs, Lm is set to “1” only for the immediately following one frame period, and after that, when Lm is incremented by one for each frame period and reaches the normal time width Lmd, the value thereafter To maintain.

  First, the adjustment amount target value generation unit 301 divides the case into four ways as in the first method, and determines the adjustment amount target value M (n).

  The adjustment amount target value averaging unit 321 averages the adjustment amount target value M (n) over the past Lm frames, and outputs an average adjustment amount target value AM. This averaging is expressed by the following equation (17).

  As the time width Lm is larger, the adjustment amount F (n) changes more slowly, and thus the parallax amount of the displayed image changes more slowly. On the other hand, the smaller the time width Lm, the faster the change in the adjustment amount F (n), and the faster the change in the parallax amount of the displayed image.

After a sufficient time has elapsed since the previous scene change was detected without a scene change (that is, after the Lmd frame period has elapsed since the previous scene change), Lm = Lmd and the adjustment amount target value The averaging unit 321 calculates an average (average adjustment amount target value) AM of the adjustment amount target value M (n) over the past Lmd frames.
In this way, by taking the time average of the adjustment amount target value M (n), a sudden change in parallax and false detection can be mitigated.

When a scene change occurs, the time width Lm is set to “1”. As a result, the output of the adjustment amount target value averaging unit 321 becomes the same value as the input. That is, averaging over a plurality of frames is not performed, and the adjustment amount target value M (n) is output from the adjustment amount target value averaging unit 321 as it is as the average adjustment amount target value AM.
Accordingly, the adjustment amount F (n) can be set to a value so that the parallax amount in the displayed image immediately becomes a value within the limit range.

After the scene change, the time width Lm is increased by 1 for each frame.
As a result, gradually return to the normal state (a state in which a sudden change in the amount of parallax for adjustment is suppressed) without being affected by the parallax before the scene change (maximum value on the pop-out side, maximum value on the pull-in side). Can do.

The operations of the adjustment amount difference generation unit 302b, the adjustment amount change generation unit 303, the adjustment amount generation unit 304, and the memory 305 are the same as those in the first method. That is, when no scene change has occurred, the adjustment amount difference generation unit 302b is based on the average adjustment amount target value AM obtained by the equation (17) as described above (as in the first method described above). , The average adjustment amount target value AM is used instead of the adjustment amount target value M (n)), and the difference between the average adjustment amount target value AM and the adjustment amount F (n−1) one frame before is set as the adjustment amount difference S. calculate. This process is expressed by the following equation (18).
S = AM-F (n-1) (18)

  The adjustment amount change generation unit 303 converts the adjustment amount difference S into the adjustment amount change K in the same manner as in the first method.

As shown in the above equation (8), the adjustment amount generation unit 304 includes the adjustment amount F (n−1) one frame before and the adjustment amount change K generated by the adjustment amount change generation unit 303. Is added to obtain the adjustment amount F (n).
The adjustment amount F (n) obtained as described above is an output of the parallax adjustment unit 3.

  In this way, the adjustment amount F (n) is limited to be within a certain range when the scene change has not occurred. By controlling in this way, a sudden change in the amount of parallax is suppressed, and the amount of parallax is adjusted slowly.

  When a scene change occurs, the scene change detection signal SC = 1 over one frame. At this time, in the switching units 326 and 307, the switching contact C is connected to the A side, and the processing by the adjustment amount difference generation unit 302b, the adjustment amount change generation unit 303, and the adjustment amount generation unit 304 is performed. Instead, as shown in the following equation (19), the average adjustment amount target value AM is immediately output as the adjustment amount F (n) as it is.

That is,
When SC = 1 F (n) = AM (19)

  Thus, when a scene change occurs, the adjustment amount target value M (n) is not averaged over a plurality of frames, and the adjustment amount target value M (n) is output as it is as the average adjustment amount target value AM. The adjustment amount F (n) is immediately changed to a value equal to the average adjustment amount target value AM, and accordingly to a value equal to the adjustment amount target value M (n).

  In this way, when a scene change occurs, the time average process over a plurality of frames and the restriction on the amount of parallax change are interrupted so that the adjustment amount F (n) immediately matches the adjustment amount target value M (n). Thus, the parallax amount can be immediately set to a value within the limit range without causing a sense of incongruity.

  When no scene change has occurred, the adjustment amount target value M (n) is time-averaged over the time width Lmd, so that it is considered that a rapid parallax change over time can be avoided, and the equation (18) The average adjustment amount target value AM may be output as it is as the adjustment amount F (n) without performing the processing of equation (8). Further, when a scene change occurs, the time width Lm in averaging according to the above equation (17) is once set to “1” and then gradually increased, so that the average adjustment amount target value AM is immediately adjusted as it is. It may be output as F (n). In this case, the adjustment amount difference generation unit 302b, the adjustment amount change generation unit 303, the adjustment amount generation unit 304, the memory 305, and the switching units 306 and 307 in the configuration of FIG. 9 may be omitted, and the configuration of FIG.

The processing in this case is represented by the following formula (20).
For both SC = 0 and SC = 1,
F (n) = AM (20)

Next, the operation of the image generation unit 4 will be described. Based on the adjustment amount F (n) output from the parallax adjustment unit 3, the image generation unit 4 generates a new left-eye image and right-eye image from the left-eye input image Da1 and the right-eye input image Db1. It produces | generates and outputs as the display image Da3 for left eyes, and the display image Db3 for right eyes.
Here, the image generation unit 4 generates an image by pairing the left-eye image and the right-eye image based on the adjustment amount F (n).

  Assume that the left-eye input image Da1 and the right-eye input image Db1 shown in FIG. Image portions B1L and B1R indicated by circles with cross-hatching in the figure correspond to the same object viewed stereoscopically. The image part B1L corresponding to the object of the left-eye input image and the image part B1R corresponding to the object of the right-eye input image originally have parallax d1.

FIG. 11B shows a state where the entire left-eye input image is translated leftward in the horizontal direction, and the entire right-eye input image is translated rightward, that is, in the horizontal range indicated by reference numerals GL2 and GR2, respectively. The state after horizontally moving a certain image so that it may become an image in the horizontal range shown by code | symbol GL3 and GR3, respectively is shown.
Referring to FIG. 11B, the parallax between the left-eye image portion B1Ls and the right-eye image portion B1Rs corresponding to the object B1 is d1s, which is smaller than the original parallax d1. In other words, it appears to the user that the projecting object has moved in the direction in which it is drawn, that is, toward the back.

  The relationship between the pixel shown on the display surface and the position of the object perceived by the user is shown in FIGS. FIG. 12A is the same as FIG. P1L in FIG. 12A is a pixel in the image portion B1L corresponding to the object of the left eye image in FIG. 11A, and P1R is an image portion B1R corresponding to the object in the right eye image of FIG. Let's represent the middle pixel. The pixel P1L for the left eye image and the pixel P1R for the left eye image have parallax d1, and the user perceives that there is an object at the point F1 in the space.

  FIG. 12B is a diagram in which the left-eye image displayed on the display surface is moved to the left and the right-eye image is moved to the right as shown in FIG. 11B. That is, in FIG. 12B, the pixel P1L in the image portion B1L corresponding to the object of the left-eye image is at the position indicated by the symbol P1Ls, and the pixel P1R in the image portion B1R corresponding to the object of the right-eye image is It moves to the position indicated by the symbol P1Rs. As shown in FIG. 12B, the parallax between the pixel P1Ls for the left eye image and the pixel P1Rs for the right eye image is d1s (<d1), and the pixel P1Ls for the left eye image and the pixel P1Rs for the right eye image. The user perceives that there is an object at the point F1s in the space. That is, it is perceived that the projecting object has been pulled in by dF inward.

Thus, the parallax can be changed by horizontally moving the left-eye input image and the right-eye input image. This movement amount dF is called a parallax offset amount.
When the parallax offset amount is increased in the negative direction, the entire left eye image is horizontally moved to the left and the entire right eye image is horizontally moved to the right.
On the other hand, when the parallax offset amount is increased in the positive direction, the entire left eye image is horizontally moved to the right, and the entire right eye image is horizontally moved to the left.
When the parallax offset amount is increased in the negative direction, the entire screen is drawn. When the parallax offset amount is increased in the positive direction, the entire screen is projected.
That is, the depth direction position of the visually recognized object can be changed by adjusting the parallax offset amount.

The image generation unit 4 receives the adjustment amount F (n) output from the parallax adjustment unit 3 and performs parallax adjustment by moving the screen in the horizontal direction using the adjustment amount F (n) as the parallax offset amount. Specifically, if F (n) is a positive number, the left eye image is moved horizontally to the right of F (n) / 2 pixels,
The right eye image is moved horizontally to the left by F (n) / 2 pixels.
On the other hand, when the adjustment amount F (n) is a negative number, the left eye image is moved horizontally to the left of F (n) / 2 pixels, and the right eye image is moved horizontally to the right of F (n) / 2 pixels. Move to.
Thus, by moving the left eye image and the right eye image in opposite directions, the parallax between the left eye image and the right eye image changes by a total of F (n) pixels.

  As described above, as a result of moving the entire image for the left eye and the entire image for the right eye in the reverse direction in the horizontal direction, the left image is displayed on the screen as shown in FIG. 11B. Portions (common portions) GLc and GRc where both the eye image and the right eye image exist, a missing portion GLm where the left eye image does not exist, and a missing portion GRm where the right eye image does not exist are generated. As a result of the above movement, a part (left end) GLa of the left eye image protrudes from the screen, and a part (right end) GRa of the right eye image protrudes from the screen.

The missing portions GLm and GRm on the screen need to be filled with some image. In this case, the image content of the right eye image portion (corresponding portion) GRb displayed at the same position as the missing portion GLm of the left eye image corresponds to the left eye image used to fill the missing portion GLm. Is desirable. Similarly, the image content of the left eye image portion (corresponding portion) GLb displayed at the same position as the missing portion GRm of the right eye image corresponds to the right eye image used to fill the missing portion GRm. Is desirable.
Therefore, in the process of filling in the missing portions GLm and GRm, a process of simultaneously replacing the corresponding portions GRb and GLb with the corresponding images is also performed.

In order to fill in the missing portions GLm, GRm and the corresponding portions GRb, GLb, a process (scaling) for extending the common portions GLc, GRc of the left and right images to the left or right is performed, or a predetermined pattern or a solid image prepared separately is used. Processing to display can be performed.
For the scaling, a method of extending the common portions GLc and GRc only in the horizontal direction or a method of extending not only in the horizontal direction but also in the vertical direction can be used, and the extension rate may be uniform over the entire screen. May be extended at different rates.

Hereinafter, scaling will be described with reference to FIGS. 13 (a) and 13 (b).
FIG. 13A shows an image after parallax adjustment, as in FIG.
In scaling, as shown in FIGS. 13A and 13B, the images of the common portions GLc and CRc located at the center are stretched left and right to create left and right portions.
Since the width of each of the missing portions GLm, GRm and the corresponding portions GLb, GRb is the same as ½ of the parallax offset amount, that is, the adjustment amount F (n), the adjustment amount is generated when the missing portion and the corresponding portion are created. The stretch rate is determined based on F (n), the image after the horizontal movement is stretched according to the stretch rate, and the portion that is located outside the screen after the stretch is removed (that is, the screen The left-eye display image Da3 and the right-eye display image Db3 are generated.
Note that the horizontal movement may be performed after the stretching is performed first.

In the example shown in FIGS. 14A and 14B, the common portion is extended in the horizontal direction so as to have the same horizontal size as the original image.
Among these, in the example shown in FIG. 14A, the stretching rate is constant. That is, in the example of FIG. 14A, the enlargement ratio obtained by dividing the horizontal size of the original image by the horizontal size of the common portion is applied to the entire screen. The horizontal size of the common part is obtained based on the adjustment amount F (n). That is, if the number of pixels in the horizontal direction of the original image is Xo and the adjustment amount represented by the number of pixels is F (n), the stretching rate Rho is
Rho = Xo / (Xo-F (n))
Given in.

  As shown in FIG. 14A, if the image is stretched at a certain stretch rate over the entire screen, the image becomes a little horizontally long, and there is no sense of incongruity.

In the example shown in FIG. 14B, the expansion rate Rh is “1” at the center of the screen, and is set to a larger value as it approaches the left and right ends. The change in the stretching rate Rh may be linear or curvilinear, and may be continuous (linear) or discontinuous (nonlinear). Furthermore, the width of the central portion where the stretching rate Rh is “1” may be zero. In other words, the stretching ratio Rh may gradually increase from the central portion toward the left and right end portions.
In this method, in the central part of the screen that is often watched by the user, the sense of incongruity can be eliminated or minimized by not changing the aspect ratio of each part of the image or by suppressing the change in the aspect ratio. On the other hand, the closer to the left and right ends, the larger the stretching ratio Rh. However, the image near the left and right ends is often not watched as much as the central portion, and it is difficult to notice that it has been stretched.

In the example shown in FIG. 14C, the image is stretched in the horizontal direction and also in the vertical direction.
The stretching ratio may be the same in the horizontal direction and the vertical direction as in the example shown in FIG. 14C, but may be different. For example, the vertical stretching rate Rv may be smaller than the horizontal stretching rate Rh.
In the example shown in FIG. 14C, the horizontal extension rate Rh is set to the same value (Rho) as in the example of FIG. 14A, and also in the vertical direction, the same extension rate Rvo (= Rho) as in the horizontal direction. ).
In this way, the aspect ratio of each part of the image can be kept the same as the original image over the entire screen.
Note that, since the upper end and lower end of the original image protrude from the screen, they are cut off, but the aspect ratio of each part of the image is maintained, which is preferable in that respect.

In the method of FIG. 14D, a predetermined pattern or plain wallpaper is displayed on the left and right missing portions and corresponding portions. In the case of a plain image, black, gray, etc. are displayed.
Further, it may represent character information different from the image in the central portion, for example, a logo or a character for advertisement.
This method is preferable in that respect because it is not necessary to stretch the image and is easy to process, and the aspect ratio of the image of the common portion can be maintained as the original image.

  The method for filling the above-described missing portion and the corresponding portion is not limited to the above-described method described with reference to FIGS. 14A to 14D, and other methods may be used.

  In this way, the image generation unit 4 moves the left-eye image and the right-eye image in the horizontal direction and moves in accordance with the adjustment amount F (n) output from the parallax adjustment unit 3. The left-eye display image Da3 and the right-eye display image Db3 are generated by filling the missing portion generated as a result of the above with any image. As a result, even if the adjustment amount F (n) changes, an image at either end of the left and right images is not lost or a different image does not appear, and a good stereoscopic image without a sense of incongruity can be obtained.

  Further, instead of determining the enlargement rate based on the adjustment amount F (n) in the image generation unit 4, the maximum value of the parallax amount adjustment amount may be predicted and extended at a predetermined extension rate. In this case, stretching may be performed before horizontal movement for parallax adjustment.

  Alternatively, the image may be displayed as it is without being stretched, and the missing portions GLm and GRm generated as a result of the horizontal movement by the parallax adjustment may remain on the black screen. In this case, the image portions GRb and GLb corresponding to the missing portions GLm and GRm are displayed as they are.

  Next, the operation of the display output unit 5 will be described. As a display method that enables stereoscopic display, a method for displaying a left-eye image and a right-eye image at spatially different coordinates, or a display of a left-eye image and a right-eye image in temporally different frames There is a method to do. The display output unit 5 receives the left-eye display image Da3 and the right-eye display image Db3 from the image generation unit 4, and outputs image signals Da4 and Db4 in a format suitable for supply to the display unit 200.

  The display output unit 5 may output the image signal after being converted into a certain format. For example, the effects of the present invention can be obtained even when a recording / reproducing apparatus including the stereoscopic image processing apparatus 100, for example, a Blu-ray reproducing / recording apparatus, is connected to a TV receiver including a display unit.

  As described above, the parallax is detected from the left-eye input image and the right-eye input image, and the left-eye input image and the right-eye input image are changed while controlling the adjustment speed based on the detected parallax. Therefore, the parallax amount range suitable for the user can be adjusted naturally without a sense of incongruity, and even when a scene change occurs, switching can be performed without a sense of incongruity as a 3D image.

Embodiment 2. FIG.
FIG. 15 is a diagram showing a configuration of a stereoscopic image processing apparatus according to Embodiment 2 of the present invention. In FIG. 15, the parallax detection unit 1, the scene change detection unit 2, the parallax adjustment unit 3, the image generation unit 4, and the display output unit 5 have the same configuration as in the first embodiment. The stereoscopic image processing apparatus 100 according to the second embodiment is different in that an image delay unit 6 is added.

  The image delay unit 6 compensates for the time required for the parallax detection by the parallax detection unit 1 and the scene change detection by the scene change detection unit 2, and generates the image after delaying the input image data Da1 and Db1 by one or more frames. This is for inputting to the part 4.

  The image delay unit 6 is composed of, for example, a frame memory. In that case, a pair of left-eye input image data Da1 and right-eye input image data Db1 (which are paired with each other) are written into the frame memory and read out from the frame memory after one or more frame periods. A pair of left-eye image data Da2 and right-eye image data Db2 are output. Thereby, a delay of one or a plurality of frame periods is performed. The reason for delaying in this way is to delay the image data and synchronize the data in accordance with the time required for parallax detection and scene change detection.

The parallax detection unit 1 receives the input image data Da1 and Db1, and detects the pop-out side maximum value Tmax and the pull-in side maximum value Hmax of one screen of the image represented by these data.
Since detection requires information for one screen (one frame), the detection result is output with a delay of one frame period or more.
When the scene change detection unit 2 obtains and detects the average luminance for one screen, the detection result is output with a delay of one frame or more.
Therefore, in the case of a configuration in which the image delay unit 6 is not provided (configuration in FIG. 1),
Compared to the input timing of the image data for the left eye and the image data for the right eye to the image generation unit 4,
The supply of the adjustment amount F (n) from the parallax adjustment unit 3 is delayed by one frame or more, and the parallax amount adjustment is delayed by one frame or more.
In particular, since the scene change detection signal SC output from the scene change detection unit 2 is delayed by one frame, the result of detecting the scene change is reflected in the amount of parallax adjustment.

  In a continuous scene, the image of the previous frame and the image of the current frame are correlated, so the amount of parallax does not change significantly between the image of the previous frame and the image of the current frame, and is generated based on the parallax of the previous frame. There is no problem even if the current frame (adjustment target frame) is adjusted with the adjusted amount. However, when a scene change occurs, the image immediately after the scene change is completely different from the previous frame image. There is a problem that the parallax adjustment is performed by the adjustment amount determined based on the parallax of the previous frame, which is different from the target frame).

  In the second embodiment, in order to solve this problem, the image delay unit 6 is provided for the input image data Da1 and Db1, and the parallax detection, the scene change detection, and the image generation unit 4 of the adjustment amount F (n) based on these are provided. And the supply of the corresponding image to the image generation unit 4 are synchronized. That is, parallax detection and scene change detection are performed based on the input image data Da1 and Db1, and an adjustment amount is determined based on these detection results so as to synchronize with the supply to the image generation unit 4. The image data Da2 and Db2 obtained by delaying the input image data Da1 and Db1 by the image delay unit 6 are supplied to the image generation unit 4.

As shown in the figure, the image delay unit 6 is also used as a circuit block in which a frame delay occurs due to scaling or other image processing in the entire image display apparatus system without preparing a separate one as shown in the figure. Good. In that case, the image data input to the circuit block is input to the parallax detection unit 1 and the scene change detection unit 2 as input image data Da1 and Db1, and the image is output from the circuit block and delayed by one or more frame periods. Data may be supplied to the image generation unit 4 as image data Da2 and Db2.
For example, in the case of performing stretching at a predetermined stretching rate, a separate scaling unit for stretching is provided in the preceding stage of the image generation unit 4, and time for processing in the scaling unit is necessary for synchronization. The delay time may be used as a part or all of the delay time.

  Based on the adjustment amount F (n) output from the parallax adjustment unit 3, the image generation unit 4 uses the left-eye image Da2 and the right-eye image Db2 delayed by the image delay unit 6 to display the left-eye display image Da3. The right-eye display image Db3 is generated and output. Here, the image generation unit 4 generates an image by pairing the left-eye image and the right-eye image based on the adjustment amount F (n).

  As described above, by using the image delay unit 6, it is possible to eliminate the shift in the timing of the parallax adjustment due to the delay associated with the parallax detection or the scene change detection, and the parallax adjustment is performed on the frame image at the optimal timing.

  Further, the parallax is detected from the left-eye input image Da1 and the right-eye input image Db1, and the left-eye image Da2 and the right-eye image that are delayed and synchronized by the image delay unit 6 based on the detected parallax. By adjusting the parallax amount of Db2 while controlling the speed of adjustment, the parallax amount range suitable for the user can be adjusted naturally without any sense of incongruity, and even when a scene change occurs, parallax adjustment can be performed at the optimal timing. Thus, it is possible to obtain a good three-dimensional image without natural discomfort.

As described above, in both cases where the amount of parallax exceeds the limit value on the pop-out side and where the amount of parallax exceeds the limit value on the pull-in side, a limit is imposed on the speed of adjustment for adjusting the amount of parallax. However, only when the amount of parallax exceeds the limit value on the pop-out side and when the amount of parallax exceeds the limit value on the pull-in side, the speed of adjusting the amount of parallax is limited. As good. For example, the speed of adjusting the amount of parallax is limited only when the amount of parallax exceeds the limit value on the pop-out side, and when the amount of parallax exceeds the limit value on the pull-in side, the speed of adjusting the amount of parallax is limited. The adjustment may be performed so that the amount of parallax is immediately within the limit value without any limitation.
Furthermore, the present invention can also be applied to a configuration in which the amount of parallax is adjusted when the amount of parallax exceeds a limit value only on one of the pop-up side and the pull-in side.

  1 parallax detection unit, 2 scene change detection unit, 3 parallax adjustment unit, 4 image generation unit, 5 display output unit, 6 image delay unit, 100 stereoscopic image processing device, 200 display unit

Claims (15)

A left-eye image represented by the left-eye input image data, the input image data for the left eye and the input image data for the right eye paired with each other. A parallax detection unit that detects a maximum value of the parallax amount between the images for the right eye represented by the input image data for the right eye or an average value over a plurality of frames of the maximum value;
A scene change detection unit for detecting a scene change in the left eye image and the left eye image;
A parallax adjustment unit that generates an adjustment amount for causing the parallax amount in the displayed image to be equal to or less than the limit value when the detected maximum value of the parallax amount exceeds a set limit value;
From the image data obtained by delaying the input image data for the left eye and the input image data for the right eye, or the input image data for the left eye and the input image data for the right eye for one or more frame periods, An image generation unit that generates display image data for the left eye and display image data for the right eye in which the parallax amount is adjusted according to the adjustment amount;
A display output unit that outputs a display image signal of a predetermined format from the display image data for the left eye and the display image data for the right eye,
The parallax adjustment unit includes:
When a scene change is detected by the scene change detection unit and the parallax amount detected by the parallax detection unit exceeds the limit value, the parallax amount in the displayed image is immediately equal to or less than the limit value. The value of the adjustment amount is determined so that
When no scene change is detected by the scene change detection unit and the amount of parallax detected by the parallax detection unit exceeds the limit value, the amount of parallax in the displayed image gradually increases to the limit value. The stereoscopic image processing apparatus characterized in that the value of the adjustment amount is determined as follows. For a frame of data,
The parallax detection unit detects that the maximum value exceeds the limit value,
When a scene change is detected by the scene change detection unit,
The parallax adjustment unit determines a value of the adjustment amount so that a parallax amount of a frame generated by the image generation unit is equal to or less than the limit value,
For a frame of data,
The parallax detection unit detects that the maximum value exceeds the limit value,
When no scene change is detected by the scene change detection unit,
The parallax adjustment unit determines a value of the adjustment amount so that a change in the adjustment amount between frames that are generated one after another by the image generation unit does not exceed a predetermined value. The stereoscopic image processing apparatus according to claim 1, wherein the amount becomes equal to or less than the limit value after one or a plurality of frames.   The stereoscopic image processing apparatus according to claim 2, wherein the predetermined value is a value that changes according to a difference between the maximum value and the limit value. The parallax adjustment unit includes:
When no scene change is detected by the scene change detection unit,
Based on the adjustment amount difference obtained by subtracting the adjustment amount one frame before from the adjustment amount target value obtained by subtracting the limit value from the maximum value, an adjustment amount change is generated,
The adjustment amount of the current frame is generated by adding the adjustment amount change amount to the adjustment amount of the previous frame,
The adjustment amount change amount is generated so as to have an absolute value smaller than the adjustment amount difference in an area where the absolute value of the adjustment amount difference is equal to or larger than a predetermined value. A three-dimensional image processing apparatus according to claim 1. The relationship between the absolute value of the adjustment amount difference and the absolute value of the adjustment amount change generated based on the adjustment difference is as follows:
When the adjustment amount difference is a positive value;
When the adjustment amount difference is a negative value,
The stereoscopic image processing apparatus according to claim 4, wherein the stereoscopic image processing apparatuses are different from each other. The parallax adjustment unit includes:
When a scene change is detected by the scene change detection unit,
The stereoscopic image processing apparatus according to any one of claims 1 to 3, wherein an adjustment amount target value obtained by subtracting the limit value from the maximum value is output as the adjustment amount. When no scene change is detected by the scene change detection unit,
The parallax adjustment unit includes:
An average value of the maximum values is obtained by averaging the maximum values over a plurality of frames,
The stereoscopic image processing apparatus according to claim 1, wherein the adjustment amount is determined based on an average value of the maximum values. When a scene change is detected by the scene change detection unit,
The parallax adjustment unit includes:
The stereoscopic image processing apparatus according to claim 7, wherein the adjustment amount is determined based on the maximum value instead of the average value of the maximum values. When no scene change is detected by the scene change detection unit,
The parallax adjustment unit determines an average adjustment amount target value by averaging the adjustment amount target value over a plurality of frames,
The stereoscopic image processing apparatus according to any one of claims 1 to 3, wherein the adjustment amount is determined based on the average adjustment amount target value. When a scene change is detected by the scene change detection unit,
The stereoscopic image processing apparatus according to claim 9, wherein the parallax adjustment unit determines the adjustment amount based on the adjustment amount target value instead of the average adjustment amount target value. As the maximum value of the parallax amount, the maximum value on the pop-out side and the maximum value on the pull-in side are obtained,
As the limit value, a limit value on the protruding side and a limit value on the drawing side are set,
When the maximum value on the popping side exceeds the limit value on the popping side, and / or when the maximum value on the pulling side exceeds the limit value on the pulling side,
The stereoscopic image processing apparatus according to claim 1, wherein the parallax adjustment unit generates an adjustment amount for adjusting a parallax amount of a displayed image. The parallax adjustment unit includes:
Adjusting the amount of parallax for the detected maximum value on the pull-in side exceeding the pull-side limit value;
The stereoscopic image processing apparatus according to claim 11, wherein the stereoscopic image processing apparatus is performed in preference to the adjustment of the parallax amount when the detected maximum value on the protruding side exceeds the limiting value on the protruding side. The scene change detection unit
The stereoscopic image processing apparatus according to claim 1, wherein the scene change is detected based on a change in the amount of parallax detected by the parallax detection unit. So as to compensate for the time for detection of parallax by the parallax detection unit and detection of scene change by the scene change detection unit,
The stereoscopic image processing apparatus according to claim 1, further comprising an image delay unit that inputs the input image data to the image generation unit after delaying the input image data by one or more frames. A stereoscopic image processing apparatus according to any one of claims 1 to 14,
A stereoscopic image display device comprising: a display unit configured to display a stereoscopic image based on the display image signal output from the display output unit of the stereoscopic image processing device.

Images