JPH05300489A - Encoding/decoding device of image signal - Google Patents

Abstract

PURPOSE:To obtain an encoding device and a decoding device capable of accurately restoring an image signal to an image without transmitting the conversion factor outside an image frame, in a conversion encoding by a non-object base. CONSTITUTION:An endpoint calculation circuit 100 and an endpoint calculation circuit 200 are provided on a coding side a decoding side, respectively. The endpoint calculation circuit 100 adds an image signal a conversion factor is symmetrical to an image endpoint to the outside of an image frame. The endpoint calculation circuit 200 prepares the signal group of the outside of the image frame by turning back the conversion factor within image frame symmetrically to the endpoint.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal encoding apparatus and an image signal decoding apparatus, and more particularly to a method for processing image end points in transform encoding having an asymmetric base such as wavelet transform.

[0002]

2. Description of the Related Art Since transform coding having an asymmetric base has not yet been put to practical use in the industrial field, it is not clear how to process image end points in the past. However, since the conversion by the base and the inverse conversion are a kind of digital filter processing, it suffices to add an appropriate signal to the outside of the image frame in encoding and transmit the conversion coefficient outside the image frame by the amount necessary for decoding. become.

FIG. 3 is a block diagram showing a general one-dimensional wavelet transform (encoding) circuit.
Reference numeral 10 is a high-pass filter (hereinafter, referred to as “high-pass filter”), 30 is a down sampler connected to the high-pass filter 10, 20 is a low-pass filter (hereinafter, “low-pass filter”), and 31 is low. A down sampler connected to the low-pass filter 20, 11 is a high-pass filter connected to the down-sampler 31, 32 is a down-sampler connected to the high-pass filter 11, 21 is a low-pass filter connected to the down-sampler 31, 33 Is a down-sampler connected to the low-pass filter 21, 12 is a high-pass filter connected to the down-sampler 33, 34 is a down-sampler connected to the high-pass filter 12, 22 is a low-pass filter connected to the down-sampler 33 , 35 are downsamplers connected to the low-pass filter 22.

FIG. 4 is a block diagram showing an inverse conversion (encoding) circuit for FIG. 3, in which 74 is an upsampler, 52 is a high-pass filter connected to the upsampler 74, 75 is an upsampler, and 62 is an upsampler. Up sampler 75
Low-pass filter connected to, 72 is an upsampler, 5
1 is a high-pass filter connected to the upsampler 72, 7
3 is an upsampler connected to the highpass filter 52 and the lowpass filter 62, 61 is a lowpass filter connected to the upsampler 73, 70 is an upsampler, 50 is a highpass filter connected to the upsampler 70, 71 Is an upsampler connected to the highpass filter 51 and the lowpass filter 61, and 60 is a lowpass filter connected to the upsampler 71.

Next, Da having an even-order asymmetric base
The operation will be described by taking the ubechies wavelet transform as an example. The order of the base is 2M, and the bases of the low band and the high band are g = (g0, g1, ..., G2M−, respectively).
1), h = (h0, h1, ..., H2M-1). Here, h0 = g2M-1, h1 = -g2M-2, ..., h
2M-1 = -g0. Also, since the basis is asymmetric, at least g0 ≠ g2M-1, g1 ≠ g2M-2,
, GM-1 ≠ gM holds. In FIG. 3, low-pass filters 20-22 and high-pass filters 10-12 are shown.
Is a digital filter having impulse responses of bases g and h, respectively.

In g and h, the basis in which the order of the elements is reversed is g '= (g2M-1, g2M-2, ..., g
0), h ′ = (h2M-1, h2M-2, ..., h0), the low-pass filters 60 to 62 and the high-pass filters 50 to 52 in FIG. 4 use g ′ and h ′ as impulse responses, respectively. It is a digital filter that has.

FIG. 5 shows the arrangement of sample points of the image signal and the transform coefficient. In the figure, Xi (i = 0, 1, 2, ...)
Is a sample point of the image signal, W2i, Y2i, V4i, U
4i, Q8i, and P8i are the sample points of the transform coefficient. Here, X0 is the end point of the image. The high-pass filter 10 and the low-pass filter 20 divide the image signal Xi into a high-pass conversion coefficient Wi and a low-pass conversion coefficient Yi. Next, the down samplers 30 and 31 thin out Wi and Yi to 2: 1 to obtain W2i and Y2i, respectively. Y2i is further divided into V4i and U4i by similar processing, and U4i is further divided into Q8.
i, P8i. From the above, downsampler 3
From 0, 32, 34 and 35, the conversion coefficient W2i,
A signal sequence of V4i, Q8i, P8i will be output. This signal sequence is usually coded after processing such as quantization.

Next, on the decoding side, the upsampler 7
0, 72, 74, 75 respectively have a conversion coefficient W2
i, V4i, Q8i, P8i signal sequences are given.
The upsamplers 74 and 75 have conversion coefficients Q8i and P8i.
0 is inserted between the signal sequences. High-pass filter 5
The 2 and low-pass filter 62 extracts a high-frequency component and a low-frequency component from the 0-inserted signal sequence. By this filtering, the signal is interpolated in the portion where 0 is inserted. By synthesizing the interpolated signals, the signal sequence of the conversion coefficient U4i is obtained at the input of the upsampler 73. The bases g and h are set so that this U4i exactly matches U4i obtained from the downsampler 33 on the code side, except for errors such as quantization distortion. Up samplers 72, 73, high-pass filter 51, low-pass filter 6
1 performs similar processing on U4i and V4i to obtain the conversion coefficient Y2i. Furthermore, the upsamplers 70, 7
1, the high-pass filter 50, the low-pass filter 60, W2i,
The original image signal Xi is restored from Y2i.

Here, the processing of the image end points will be described by taking the conversion / inverse conversion from Xi to W2i, Y2i as an example. On the decoding side, the W2i (i =
0, 1, 2, ...) is given, and it is assumed that a signal sequence Wi having 0s inserted between them is obtained at its output, as shown in FIG. Here, Wi = 0 (i = odd number).

Further, the high-pass filter 51 and the low-pass filter 6
Y2i is output from 1 and the upsampler 71
It is assumed that the 0-interpolated signal sequence Yi is obtained. At this time, in the high-pass filter 50, in order to filter the signal W0, m reference points w1 and w
2, ..., Wm (m <2M) are required. Here, m is an integer uniquely determined by the shape of the basis. Similarly, the low-pass filter 60 has 2M-m-1 reference points y1, y2, ..., Y2M-m outside the image frame for filtering Y0.
-1 is required. Since Xi is obtained by synthesizing the filtered signals, the image near the end point cannot be accurately restored when there is no conversion coefficient outside the image frame.

Next, the processing on the code side for calculating the conversion coefficients wi and yi outside the image frame will be described. In order to calculate the transform coefficient wm in the high-pass filter 10, 2M-m-1 reference points xm + 1 and xm + located further outside the transform coefficient wm.
2, ..., X2M-1 are required. Further, in order to calculate the conversion coefficient y2M-m-1 in the low-pass filter 20, m reference points x2M-m, x2M-m + outside thereof are calculated.
, ..., x2M-1 are required. That is, in any case, it is necessary to add 2M-1 sample points as reference points outside the image frame.

With the conversion coefficients V4i, Q8i, and P8i, the calculation range outside the image frame is further increased.

[Problems to be Solved by the Invention]

In the conventional coding apparatus, since the end point processing is performed as described above, it is necessary to transmit the transform coefficient outside the picture frame, which causes a problem that the coding efficiency is deteriorated. Furthermore, on the code side, it is necessary to determine a sample value outside the image frame. At this time, if image discontinuity occurs at the end points, the conversion coefficient in the high frequency band becomes a large value, and the code generation amount further increases. There was a problem. Especially, it is a very undesirable property that the code amount increases in the corner of the image which is not so important visually.

The present invention has been made to solve the above problems, and an object of the present invention is to obtain an encoding / decoding device that does not need to transmit transform coefficients outside the image frame.

[0015]

An encoding apparatus according to the present invention is such that an image signal is expanded outside an image frame and a transform coefficient is calculated using this signal.

Further, the decoding apparatus according to the present invention is such that the transform coefficient is expanded outside the image frame and the image signal is calculated using this signal.

[0017]

The coding apparatus according to the present invention extends the image signal so that the transform coefficient is symmetrical with respect to the image end point.

Further, the decoding apparatus according to the present invention expands the transform coefficient by folding the transform coefficient symmetrically with respect to the end points.

[0019]

EXAMPLES Example 1. An embodiment of the present invention will be described below. FIG. 1 is a block circuit diagram of the encoding apparatus. High-pass filters 10-12, low-pass filters 20-22, and down samplers 30-35 are the same as those shown in FIG.
00 is an end point arithmetic circuit. The input signal and the outputs of the down samplers 31 and 33 are passed through the end point calculation circuit 100 to a high band filter 10 and a low band filter 20, a high band filter 11 and a low band filter 21, a high band filter 12 and a low band filter 22, respectively. Is entered.

Next, the operation will be described. As described above, in the conversion between the image signal Xi and the conversion coefficients W2i and Y2i, the conversion coefficient outside the image frame is w in order to restore the image end point.
Required up to m and y2M-m-1. Further, in order to obtain these coefficients, it is necessary to add an image outside the frame up to x2M-1 on the code side. However, the conversion factor W
The signal sequences of i, wi and Yi, yi are the end points W, respectively.
If they are symmetrical with respect to 0 and Y0, they can be correctly reproduced on the decoding side without sending conversion coefficients outside the image frame. End point calculation circuit 1
Reference numeral 00 is a circuit for adding a signal outside the image frame so that the conversion coefficients are symmetrical.

The signal value outside the picture frame is obtained as follows. First, sample values outside the frame are set to x1, x2, ..., x2
Let it be M-1. From the condition of the transformation coefficient symmetry, the following simultaneous equations are obtained.

[0022]

[Equation 1]

The expression 1 is a linear expression relating to xi, and the number of variables and the number of expressions are both 2M-1. If this is solved for xi, the sample value outside the image frame is uniquely determined.

The end point arithmetic circuit 100 adds a signal outside the image frame obtained by the equation 1 to the input image signal to create a new signal sequence, which is then processed by the high-pass filter 10 and the low-pass filter 20.
Give to. The high-pass filter 10 and the low-pass filter 20 filter this signal sequence. The conversion coefficients obtained at this time are symmetrical with respect to the end points. Down sampler 3
0 and 31 thin out the conversion coefficient signal series at a ratio of 2: 1. The output of the downsampler 30 is encoded as W2i,
At this time, it is not necessary to encode the transform coefficient outside the image frame. The output of the down sampler 31 is returned to the end point calculation circuit 100 again.

The end point arithmetic circuit 100 may be realized by hardware using a read only memory (ROM) or the like, or may be realized by software using a digital signal processor (DSP) or the like. In addition, although Equation 1 is a condition that all the transform coefficients are symmetric, x is calculated only from the condition regarding the transform coefficient after down-sampling that is actually transmitted.
You may decide i. In this case, the number of equations is smaller than the number of variables, so that there is a degree of freedom in how to determine xi.

In the above description, the input is an image signal, but the input may be a transform coefficient. In FIG. 1, the input to the high-pass filter 11 and the low-pass filter 21 is a signal sequence obtained by end-point processing the conversion coefficient of the previous stage by the end-point arithmetic circuit 100. The end point arithmetic circuit 100 deletes the signal outside the image frame of the conversion coefficient Y2i output from the down sampler 31,
Create another signal sequence by adding another reference point. This reference point is at least the conversion coefficient V4i, which is output from the downsamplers 32 and 33, as in the first embodiment.
It is obtained from the condition that U4i is symmetric with respect to the end point.
Of the signal series of the transform coefficient V4i output from the down sampler 32, the part outside the image frame does not need to be encoded.

The output of the down sampler 33 is returned to the end point calculation circuit 100 again and is similarly processed for the end point. Conversion coefficients Q8i and P8i obtained from the down samplers 34 and 35
Is still symmetric with respect to the end points, so signals outside the frame are not encoded.

FIG. 2 is a block circuit diagram of the decoding apparatus according to the first embodiment. In FIG. 2, high-pass filters 50 to 52,
The low-pass filters 60 to 62 and the upsamplers 70 to 75 are the same as those shown in FIG. 4, and 200 is an end point arithmetic circuit. Each conversion coefficient is given to the upsamplers 70, 72, 74 and 75 via the end point calculation circuit 200. The outputs of the high-pass filter 52 and the low-pass filter 62, and the outputs of the high-pass filter 51 and the low-pass filter 61 are the end point calculation circuit 20.
0 to the up-samplers 73 and 71, respectively.

The conversion coefficients Q8i and P8i are end-point processed by the end-point arithmetic circuit 200 before being given to the upsamplers 74 and 75. Since the conversion coefficient is calculated so as to be symmetric with respect to the end point, the end point arithmetic circuit 200 folds and adds the signal inside the image frame to the outside of the image frame. The upsamplers 74 and 75 insert 0 into this signal sequence. The high-pass filter 52 and the low-pass filter 62 filter the signal sequences in which 0s have been inserted and then combine them to obtain the transform coefficient U4i. U4
i is returned to the end point arithmetic circuit 200 again. The end point calculation circuit 200 removes the signal outside the image frame of U4i, and instead returns the signal inside the image frame to obtain a symmetrical signal sequence. Further, the same end point processing as Q8i and P8i is performed on the conversion coefficient V4i. The upsamplers 72, 73, the high-pass filter 51, and the low-pass filter 61 are U4i, V subjected to end point processing.
The conversion coefficient Y2i is restored from 4i.

In the above description, the restored signal is the transform coefficient. However, the decoding device may restore the image signal. The upsamplers 70 and 71 perform 0 insertion on the signal series of the conversion coefficients Y2i and W2i that have been end-point processed by the end-point arithmetic circuit 200. The high-pass filter 50 and the low-pass filter 60 restore the original image signal Xi by filtering the 0-inserted signal series.

It should be noted that all the above discussions can be easily applied even when the order of the base is odd.

[0032]

As described above, according to the present invention, since the signal outside the image frame is added so that the conversion coefficient becomes symmetrical with respect to the image end point, it is not necessary to transmit the conversion coefficient outside the image frame. There is an effect that an efficient encoding / decoding device can be obtained.

[Brief description of drawings]

FIG. 1 is a block circuit diagram of an encoding device according to a first embodiment of the present invention.

FIG. 2 is a block circuit diagram of the decoding device according to the first embodiment.

FIG. 3 is a block circuit diagram of a conventional encoding device.

FIG. 4 is a block circuit diagram of a conventional decoding device.

FIG. 5 is a diagram showing a signal sequence of an image signal and a conversion coefficient.

FIG. 6 is a diagram illustrating signal processing of conversion / inverse conversion.

[Explanation of symbols]

 10-12 High-pass filter 20-22 Low-pass filter 30-35 Down sampler 50-52 High-pass filter 60-62 Low-pass filter 70-75 Up-sampler 100 Code side endpoint arithmetic circuit 200 Decoding side endpoint Arithmetic circuit

Claims (4)

[Claims] 1. An image signal encoding device for converting an image signal by an asymmetric base and transmitting a signal sequence (transform coefficient) after conversion, wherein the transform coefficient is symmetric with respect to the end point at an end point of the image. An image signal coding apparatus comprising means for adding an image signal in such a manner that the image signal is added to the image signal, and performing conversion by a base using the added image signal, and then transmitting only the conversion coefficient within the image frame. .. 2. The image signal coding apparatus according to claim 1, wherein the image signal is a transform coefficient that has already been transformed. 3. A decoding device for an image signal, which restores an original image signal from a transform coefficient coded by the image signal coding device according to claim 1, and which is transmitted at an end point of the image. An apparatus for decoding an image signal, comprising means for symmetrically folding the generated transform coefficient with respect to an end point, and performing inverse transformation by a base using the folded back transform coefficient to restore an image near the end point. 4. The image signal decoding apparatus according to claim 3, wherein the restored image signal is a transform coefficient that has already been transformed.