JPH03187693A - Encoder - Google Patents

Abstract

PURPOSE:To realize encoding with high efficiency by performing scan in an oblique direction when the DPCM(Differential Pulse Code Modulation) encoding of a sub-sampled sampled value is performed. CONSTITUTION:A pre-filter circuit 102, a subsampling circuit 103, a scan conversion circuit 104, and a DPCM encoder circuit 105 are provided, and the calculation of a predictor is performed by using a neighboring sampled value including the sampled value in a diagonal direction out of the sampled values after thinning, and encoding processing is progressed sequentially in the diagonal direction. Therefore, it is possible to use a preceding sampled value in the oblique direction instead ef the preceding sampled value in a horizontal direction in which distance between the sampled values is separated after thinning as the predictor, and also, band limitation by pre-filtering is applied in the oblique direction to prevent the folding of subsampling, and inter-sampled value correlation is heightened. In such a way, the DPCM encoding with high predictive efficiency can be realized.

Description

[Detailed description of the invention] [Industrial use! ! ] The present invention relates to an encoding device that compresses and encodes information such as an image signal, and records or transmits the encoded information.

[Prior Art] When image information, audio information, etc. is digitized and transmitted, encoding is performed to reduce the amount of information to be transmitted.

One of the encoding methods is differential encoding (Differential encoding).
There is a pulse code modulation (hereinafter referred to as DPCM), which compresses the amount of information by utilizing the correlation between adjacent sample values. Specifically, a method is to decode an encoded sample value, use that decoded value to obtain a predicted value for the next sample value, and then quantize the difference between this predicted value and the input sample value and encode it. is taken.

On the other hand, thinning processing (subsampling) using visual characteristics is known as a means of data compression. This takes advantage of the fact that human visual resolution is lower in diagonal directions, and reduces the amount of data by thinning out pixels accordingly. Typical thinning patterns include a field offset structure in which data is thinned out in a staggered manner, a line offset structure in which the thinning phase is reversed every 1 lines, and the like. The above two compression methods can be used together, and is an image encoding method that performs DPCM encoding on subsampled pixels, and when decoding, interpolates thinned out pixels using the DPCM-decoded pixels to reproduce one image. is also known.

[Issue 1g to be solved by the invention] In the conventional encoding device as described above, when subsampling, which is thinning processing, is used in combination with DPCM, 1M
The distance between pixel sample points in the horizontal direction after J-subtraction processing is longer than that of the original pixel, and the inter-pixel correlation is low, so the encoding efficiency is low in DPCM using previous value prediction using conventional horizontal scanning. There was a drawback that the image quality deteriorated and the image quality deteriorated.

The present invention has been made to solve this problem, and it is an object of the present invention to provide an encoding device that has high encoding efficiency and can prevent image quality deterioration.

[Means for Solving the Problems] In order to achieve the above object, the encoding device of the present invention calculates predicted values using neighboring sample values including St tree values in the diagonal direction among the sample values after thinning. and means for sequentially advancing the encoding process in the diagonal direction.

[Operation] According to the present invention, by performing DPCM encoding while performing diagonal scanning on sample values after thinning processing, even if a device with a simple configuration uses previous value prediction, the sample values after thinning The diagonal pre-sample value can be used as a predicted value instead of the horizontal pre-sample value, which is far apart, and band limiting is performed in the diagonal direction by brisfiltering to prevent aliasing of subsampling. , since the sample value open correlation is high, DPCM encoding with high prediction efficiency can be realized. Furthermore, since filtering for interpolation of thinned sample values is performed during decoding, the encoding noise in the creation direction that occurs during DPCM encoding is smoothed, so image signals are compressed with less visual deterioration. Encoding can be realized.

[Example] Hereinafter, the present invention will be explained in detail with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an encoding device that is an embodiment of the present invention. In FIG. 1, terminal 101
The image signal supplied to the sub-sampling circuit 102 performs spatial filtering to prevent aliasing noise from occurring when pixels are thinned out by the sub-sampling circuit 103.

Figure 3 shows an example of filtering when thinning out a staggered structure.
is the vertical sampling frequency), and the shaded area in figure (a) represents the passband of the original signal, which is rectangular sampling, in the spatial frequency domain, and figure (b) shows the high frequency 3 is a diagram showing a pass band of a signal filtered by a bris filter circuit 102 to prevent component aliasing noise. FIG. As is clear from FIG. 3(b), the band in the diagonal direction is more concentrated than that in FIG. 3(a), but it is known that this difference does not cause significant deterioration due to the visual characteristics of one person. In this way, since the band in the diagonal direction is narrower than in the horizontal and vertical directions, as a result, the output of the bris filter circuit 102 has a stronger inter-pixel correlation in the diagonal direction than in the horizontal and vertical directions. Sub sample circuit 10
3, the band-limited image signal is subjected to thinning processing as shown in FIG. 4, for example. The thinned image signal is processed by the scan conversion circuit 104, for example, as shown in FIG.
As shown in Figure (b), the scan is converted in the diagonal direction. Here, as is clear from FIG. 5, the sub-sample circuit 10
The thinning process of 3 can also be realized by diagonal interlaced scanning in the scan conversion circuit 104. The DPCM encoding circuit 105 performs DPCM encoding and supplies the encoded output to a terminal 106.

Here, the operation when using the previous prediction DPCM as the DPCM encoding method will be explained using FIG.
In PCM, in order to prevent integration of quantization errors during decoding,
Although the predicted value is calculated not from the input pixel value itself but from the value decoded by the local decoder in the encoder, here, to simplify the explanation, it is assumed that prediction is performed from the previous pixel value itself.

Now, if the predicted value of pixel B in FIG. 6 is ``, then according to the present invention, as is clear from FIG. 9, 100 = αA (α is the prediction coefficient), compared to 100 = αC in the case of horizontal scanning. Since the distance between each pixel is short, high prediction efficiency can be achieved.The effect of the above-mentioned bris filter also contributes to this.

FIG. 2 is a block diagram showing a schematic configuration of an image decoding apparatus for the embodiment shown in FIG. In Figure 2, terminal 2
The encoded input supplied to 01 is:

DPCM decoded by the DPCM decoding circuit 202,
After conversion from diagonal scanning to normal scanning by the scanning inverse conversion circuit 203, the thinned out pixel values are interpolated by the interpolation circuit 204. This is done by filtering the image signal with the characteristics shown in Figure 3(b).
Normally, encoding errors in DPCM occur in areas where the difference value is large due to nonlinear quantization, but due to the characteristics of this interpolation filter, in the present invention, coding errors occur in areas where the difference value in the diagonal direction is large, that is, in the area where the diagonal spatial frequency is high. Another advantage of the present invention is that it is possible to visually suppress errors caused by errors.

[Effects of the Invention] As explained above, according to the present invention, when performing DPCM encoding on subsampled sample values, by scanning in a diagonal direction, the distance between sample values can be adjusted even in previous value prediction. By limiting the band in the diagonal direction, a value with a high sample value open correlation can be used for prediction, and highly efficient encoding can be realized. In addition, by using a filter operation with diagonal band-limiting characteristics for interpolation of thinned sample values during decoding, DPC that occurs in the area of high spatial frequency in the diagonal direction with the wooden sword method is eliminated.
For example, in the case of an image signal, it is also possible to make the M encoding error visually unnoticeable by smoothing it.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an encoding device that is an embodiment of the present invention, FIG. 2 is a block diagram showing a schematic configuration of an image decoding device for the embodiment of FIG. 1, and FIG. This figure shows an example of filtering when thinning out a staggered structure (where hs is the horizontal sampling frequency, and ■ is the vertical sampling frequency). A diagram showing the passband of a signal in the spatial frequency domain. Figure (b) shows the buri filter circuit 1 used to prevent aliasing noise of high frequency components during staggered sampling.
Figure 4 shows an example of the pixel thinning pattern in the sub-sample circuit, and Figures 5 (a) and (b) show the scanning direction by scan conversion. FIG. 6 is a diagram for explaining the previous value prediction DPCM according to the present invention. In the figure. IO2: Buri filter circuit 103: Sub-sample circuit 104: Scan conversion circuit 105: DPCM encoding circuit 202: DPCM decoding circuit 203: Scan conversion circuit 204/2-decimation interpolation circuit

Claims (1)

[Claims] An encoding device that predicts an input sample value from a sample value subjected to subsampling, quantizes and transmits a difference value between the predicted value and the input sample value, and performs the calculation of the predicted value based on the sample value after thinning. An encoding device comprising means for performing the encoding process using neighboring sample values including sample values in the diagonal direction, and means for sequentially advancing the encoding process in the diagonal direction.