JP2009100433A - Method of interpolating image frame - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method of interpolating an image frame by which burden on an interpolation circuit using a motion vector or the like that is complicated and has a risk of malfunction is reduced, an effect equal to or more than that of this method can be achieved, and the effect is quite great in practical use. <P>SOLUTION: An image of a frame to be interpolated is a spatial low resolution image. Accordingly, the objects (reduction of flicker, prevention of image quality deterioration caused by erroneous vector search) can be achieved. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

Detailed description of the invention

  The present invention relates to a method of doubling the number of frames by interpolating television frames (image frames).

  Recently, devices that interpolate and display frames have been widely developed and released. For example, LCD TVs are blurred in the moving area, so a new frame is inserted between two existing frames that are continuous in time and the number of frames is doubled. The purpose is to increase brightness and prevent flicker. With regard to the former (brightness enhancement), the brightness decreases significantly if the percentage of time displayed (hold rate) is reduced to prevent blur in the moving area. To prevent this, double the number of frames. The latter (flicker) is not a problem when the field frequency is 60 Hz as in the NTSC system, but may occur in a bright area with a large area.

  In addition, flicker is conspicuous when the field frequency is 50 Hz as in European PAL system television, so the number of frames is doubled and displayed in the same way.

Here, for simplicity, the case where the input image signal is a sequential scan is described first. In the case of interlaced scanning, signals converted to sequential scanning are handled.
In the prior art, for interpolation, a motion vector is searched from existing frames before and after the frame to be interpolated, this is time-divided, and the moving object is moved to form an interpolation frame. As a result, the number of frames per second is doubled. In some cases, the number of frames is tripled or quadrupled. Here, these are collectively referred to as doubling.

  FIG. 3 is a diagram showing the time relationship between the prior art and the signal of the present invention. 1, 2, 3, 4 indicate the frame number of the TV signal. When displaying at double speed, the time of 1, 2, 3, etc. is halved, and (01) is created from frames interpolated from two consecutive frames, for example, frames 1 and 2. In this way, (12), (23), etc. are made and arranged as shown below in the figure. Such a signal is displayed at double speed.

  FIG. 4 is a block diagram of a known apparatus for realizing this. The frame interpolation circuit 11 creates an interpolation frame from the signal delayed through the input signal 13 and the frame memory 14. Although not described in detail here, for example, a motion vector is searched from two frames 1 and 2, and based on this, an intermediate frame 12 is created.

  These frame signals and interpolation signals are added to the time shortening circuits 15 and 16 and read out in half the time. For example, the circuit 15 reads out the frame in half time while storing the frames 1, 2,. The circuit 16 reads out the signal in half time while storing the interpolated signals (12) (23).

  The switch 17 sequentially switches and outputs the outputs of these circuits 15 and 16 for display.

Problems to be solved by the invention

  As is generally known, the motion vector search described above requires extremely fast processing, and there are also erroneous searches. In interframe coding such as so-called MPEG, even if there is a certain amount of erroneous search, the coding efficiency is not significantly affected. Therefore, we would like to avoid the severe interpolation described above if it can be avoided.

  In relation to interpolation, when a TV signal is analyzed in the spatio-temporal frequency domain, it is found that the interpolation frame does not necessarily have a high resolution, as will be described later. Rather, it can be seen that the overall image quality of the low-resolution image is better than that of the high-resolution interpolation frame, which has the risk of erroneous search.

  In other words, if there is a false search in motion vector search, the image quality may be greatly degraded due to false interpolation, but if there is a low-pass filter, many will be out of the band and expected to be inconspicuous. In other words, the required accuracy of the interpolation circuit 11 including the motion search circuit is remarkably relaxed and simplified accordingly.

Here, the reason why the interpolated frame should be low resolution is explained. Assuming that the frame rate of the PAL method is assumed as a numerical value and a low resolution frame is interpolated, the signal after interpolation is
High-resolution component repetition frequency: 50 Hz,
Low-resolution component repetition frequency: 100 Hz
It becomes. Strictly speaking, the original signal in the interlaced scan is called a field, not a frame.

  As for human visual characteristics, even a fine pattern can be seen for still images, but only a rough pattern can be seen as the movement increases or the number of blinks increases. For changes in brightness of about 50 Hz, flicker is not felt unless the pattern is rough. I don't feel it blinking more than that.

  For simplicity, the spatio-temporal frequency spectrum for the interpolated TV signal for still images is shown in Fig. 2 in the [spatial frequency μ-temporal frequency f] region. In the figure, the horizontal axis is the spatial frequency axis (eg horizontal frequency μ). The vertical axis is the time frequency f.

  As described above, since the repetition frequency of the high resolution component is 50 Hz and the repetition frequency of the low resolution component is 100 Hz, the spectrum is as shown in FIG.

  The diamond-shaped square system in Fig. 2 shows the visible range, that is, the pass band of the visual system. As described above, even a fine pattern can be seen for a still image, but only a rough pattern can be seen when the movement is large, and a flicker of about 50 Hz does not feel that it is not a rough pattern. As a result, the spatio-temporal frequency band that humans can perceive is inside the diamond type as shown in the figure as a first order approximation.

  At this time, if there is no spatial low-frequency component in the component at 50 Hz, the television component does not enter the diamond-type pass characteristic. As a result, it cannot be detected visually, so it does not cause flicker. As can be seen from this, the purpose of frame interpolation can be achieved by interpolating only spatial low-frequency components.

  Since the high frequency component is 50 Hz and half of the low frequency component 100 Hz, it is desirable to compensate for the amplitude. That is, like the above, it is desirable to enhance the spatial high-frequency component of the high-resolution image twice by a known method.

  Incidentally, as is well known, in the moving image area of the image, the spectrum drawn horizontally in Fig. 2 is slanted. The position of μ = 0 does not change. In addition, at 50Hz, there is a hole in the same way. For this reason, the above operation works in the same way for this movie, so the explanation is omitted.

  If this idea is developed, it can be seen that as a method of interpolation, a spatial low-resolution image of the average value of the previous and next two frames may be used instead of shifting the moving region using motion vectors.

  Furthermore, if a slight image quality degradation is allowed, it may be possible to filter the previous frame image as it is. In this case, the circuit is further simplified such that the frame 25 in FIG.

  The case where the progressive scanning signal is used has been described above, but the same holds true for the interlaced scanning signal when the average of two frames (field) or the previous frame (field) is used. Here, strictly speaking, a frame is a field. In the interlaced scanning signal, the skipped scanning line displays black.

  The specific operation in the embodiment will be described with reference to FIG. 1, focusing on the difference from the conventional example in FIG.

  As mentioned above, the interpolated image need only have low frequency components spatially. Therefore, a low-pass filter 19 is placed between the interpolation circuit 18 and the time shortening circuit 20. The spatial low-pass filter calculates the weighted sum of the values of the relevant central pixel and the surrounding pixels, and is not described in detail because it is well known. The order of the low-pass filter 19 and the time reduction circuit 20 may be reversed.

  As mentioned above, the high frequency component emphasizes the spatial high frequency component twice. An enhancement circuit 22 is placed in front of the time reduction circuit 20. Since this configuration is well known as a filter, it will not be described in detail. The order of the high frequency emphasis circuit 21 and the time shortening circuit 20 may be reversed.

  Then, with the switch 24, these outputs are switched as before and sent to the display device. The rest is the same as the conventional example.

The invention's effect

  As described above, according to the present invention, it is possible to reduce the burden of an interpolation circuit using a motion vector which is complicated and has a risk of malfunction, and can achieve an effect equal to or better than the frame interpolation by this method. Because it is possible, it is very effective for practical use.

It is a block diagram of the present invention. It is explanatory drawing in this invention. It is explanatory drawing of a prior art example and this invention. It is a block diagram of a conventional example.

Explanation of symbols

11, 18, Frame interpolation circuit 14, Frame memory 15, 16, 21, 22, Time shortening circuit 18, Frame interpolation circuit 19, Low-pass filter 22, High-frequency enhancement circuit 25, Frame memory

Claims (1)

  An image frame interpolation method characterized by interpolating spatially low resolution images in a signal conversion device that doubles the number of frames by interpolating TV frames (image frames).

Images