JPH01135265A - Image encoder - Google Patents

Abstract

PURPOSE:To efficiently reproduce a heading image by dividing image data into plural subblocks, encoding only the converting coefficient of a reference subblock among them and, for others, encoding only the portions of the errors of the converting coefficients. CONSTITUTION:An image to be encoded is divided into blocks from an image memory 1 and read by a block reading part 2. The data are divided and composed into the subblocks by a subblock reading part 3. The subblocks are orthogonally transformed by an orthogonal transforming part 4. Next, the transformation coefficients of the subblock group are compressed by a transformation coefficient compressing part 5. At the compressing part 5, the transformation coefficients of the reference subblock obtained form the transforming part 4 are sent to a quantizing part 6. For the transformation coefficients of the other subblocks, differences from the corresponding coefficients of the reference subblock are successively obtained for 1 picture element, and only the error data are sent to the quantizing part 6. In such a way, the coefficients compressed by the compressing part 5 are quantized. The coefficients quantized by the quantizing part 6 are encoded by an encoding part 7.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a highly efficient encoding device that compresses digital image data and utilizes it for communications and file systems.

Prior Art 2 -\-7 In recent years, storage media such as magnetic disks and optical disks have become larger in capacity and displays have become more precise, and along with this, the demand for filing image data has increased. In terms of filing binary images, document filing devices have already become popular. However, in the case of multivalued image data, although the storage capacity of storage media has increased dramatically, the amount of data is large and it is difficult to put it into practical use as a file device unless the image data is compressed. Regarding the compression of multivalued image data, MU was a ``bandwidth compression'' that tried to compress the transmission frequency in an analog manner, but as technology such as integrated circuits progressed, it became possible to quantize the signal and have n-bit gradation. It is converted into digital image data and compressed.As one example, focusing on the fact that there is a correlation in images, the image data is divided into blocks, orthogonal transformation is applied, and the transformation coefficients are quantized. , there is orthogonal transform encoding (Japanese Unexamined Patent Publication No. 123280/1983).

This will be explained below using figures. FIG. 7 shows a conventional example. In FIG. 7, 71 is an image memory section, 723x7 is a block readout section that reads out nXn (n: integer) pixel data from the image memory, 73 is an orthogonal transformation section that orthogonally transforms the block data, and 74 is a block readout section that reads out nXn (n: integer) pixel data from the image memory. A quantizer 75 quantizes the transform coefficients, an encoder 75 encodes the quantized coefficients, and 76 encodes data. Image data is read out block by block from the image memory 71 by the block readout unit 72, and the block data is orthogonally transformed by the orthogonal transformation unit 73. In the literature referred to here, discrete cosine transformation (hereinafter referred to as DCT) is used. (calling) is applied. Furthermore, the transform coefficients are stored in the quantization unit 74° encoding unit 76.
quantized and encoded. In the conventional example, when reproducing an image, the reverse of the encoding is followed, the encoded data is decoded, the transform coefficients are regenerated by inverse quantization, and the Glock of the corresponding image data is regenerated from the transform coefficients. There is. In addition, when applying the conventional conversion encoding device shown here to an image file and using a headline image for image retrieval, for example, in order to create the headline image, all of the image data is reproduced, and then reduced and the headline image is used. It is necessary to create an image or create a separate index image and encode it.

Problems to be Solved by the Invention As described above, in the conventional conversion encoding device, when creating a headline image in an image file, calculation efficiency is reduced by reconstructing from the original image or creating another image. Another problem is that it is not good in terms of memory efficiency.

Means for Solving the Problems In the present invention, in order to solve the above problems, image data is divided into blocks, and each block is further divided into a plurality of subblocks obtained by subsampling different pixel positions, #J subblock is orthogonally transformed, and the obtained transform coefficients are encoded.Among the plurality of subblocks, the transform coefficient of the first subblock is encoded as is, and the first subblock is For the sub-blocks to be removed, the error between the transform coefficients of the sub-blocks and the transform coefficients of the sub-block of @whistle 1 is encoded.

Effect 5 By using the above method, it is possible to efficiently reproduce a reduced image as a heading image by reproducing a part of the sub-blocks that make up the block, and further by reproducing the remaining sub-blocks. Full images can also be obtained efficiently.

Example Hereinafter, the present invention will be explained in detail using the drawings.

FIG. 1 is a block diagram of recording and encoding image data according to an embodiment of the present invention. In the figure, 1 is an image memory section, 2 is a block reading section, 3 is a sub-block reading section, and 4 is a block diagram for recording and encoding image data. The orthogonal transform section, 5 is a transform coefficient compression section, 6 is a quantization section, 7 is an encoding section, and 8 is encoded data. FIG. 2 is a block diagram of image data reproduction/decoding. In the figure, 8 is encoded data, 9 is a decoding unit, 1o is an inverse quantization unit, 11 is a transform coefficient expansion unit, 12 is an inverse orthogonal transform unit, 1
3 is an image memory. FIG. 3a shows a detailed block diagram of the transform coefficient compression section, and FIG. 3 shows a detailed block diagram of the transform coefficient expansion section. Figure 4 shows one of the sub-samples in the present invention.
For example, in the same figure, 61\-a indicates the original image, and b to e each indicate a sub-block. FIG. 6 shows an example of the encoded data structure of the present invention. In the figure, a indicates encoded data for each subblock of one block, and b indicates encoded data for one image. 1st
In the figure, an image to be encoded is divided into blocks from an image memory 1 and read out by a block reading section 2. The data is shown in FIG. 4a. In the same figure, x,
y indicates the position of a pixel within each block.

In the figure, ○ marks represent pixels, and a group of pixels with the same number inside the ○ constitutes a sub-block. The data of the block is divided and configured into sub-blocks by the sub-block reading section 3.

Its configuration is shown in FIG. In the same figure,
The pixel configurations of the sub-blocks are shown in b to e, and are composed of pixels obtained by sub-sampling the block data every other pixel. By configuring subblocks like this,
The image data of each sub-block becomes similar. The orthogonal transform unit 4 orthogonally transforms the sub-block. Image Data Transformation 7 In Beige transformation, discrete cosine transformation is often applied, and this case is also used as an example.

Since the sub-blocks are similar image data, their transform coefficients are also similar. Taking advantage of its characteristics,
A transform coefficient compression unit 5 performs compression processing on the transform coefficients of the sub-block group. The details are shown in FIG. 3a. In the same figure, 3
1 is a transform coefficient of a sub-block, 32 is a memory for temporarily storing a reference sub-block transform coefficient, 33 is a subtraction unit, and 34 is a result of the subtraction, and the memory 32 is set to an initial value ◯ at the start of block conversion. The orthogonal transform unit 4 first obtains the transform coefficients of the reference sub-block when the initial value of the memory 32 is O.
Therefore, even if the subtraction unit 33 subtracts, that! The coefficient of M is sent to the quantization section as result 34.

At that time, the transform coefficients of the reference sub-block are stored in the memory 32. The difference between the transform coefficients of the next sub-block and the corresponding transform coefficient of the reference sub-block is taken pixel by pixel by the subtracting section 33, and as a result, only the error data is sent to the quantization section. In this way, the transform coefficients other than the reference sub-block are only the error with the corresponding transform coefficient of the reference sub-block, and since the transform coefficients of the sub-blocks are similar, the error is small, and the coefficient will be compressed. In this way, the coefficient compression unit 5
The compressed coefficients are quantized by the quantizer 6. When quantizing, the number of quantization bits is reduced for coefficients sent to the quantization unit 6 whose values are small.
Quantize by allocating more bits to larger items. For example, a large number of bits are allocated to the transform coefficients of the reference sub-block, and a small number of bits are allocated to the error coefficients of other sub-blocks since they have small values. The coefficients quantized by the quantizer 6 are converted into encoded data by the encoder 7. FIG. 6a shows an example of the encoded data structure of one block, and FIG. 6b shows an example of the encoded data structure of one image data.

In the same figure, A indicates the encoded data of the transform coefficients of the reference sub-block in Fig. 2, and B-D indicate the encoded data of the transform coefficients of the sub-block and the reference sub-block, which are not shown in Fig. 2 C-9. This is encoded data of error coefficients. This way you get 9
The encoded data is communicated or recorded in memory.

In this case, as shown in FIG. 6, there is an example in which a group of encoded data of transform coefficients of a reference sub-block is first transmitted or recorded. We will explain how to reproduce or receive such data and reproduce images. FIG. 2 shows a block diagram for reproducing and decoding image data. In the same figure,
The encoded data 8 is decoded by the decoding section 9, and the transform coefficients and error coefficients are reproduced by the inverse quantization section 1°.

The reproduced transform coefficients and error coefficients are expanded into the original coefficients by a transform coefficient expansion section 11.

A detailed block diagram of the transform coefficient expansion section 11 is shown in FIG.

In the figure, 35 indicates the reproduced transform coefficients and error coefficients, 36 a memory for temporarily storing the reference sub-block transform coefficients, 37 an adder, and 38 the addition result. The initial value is 0. Since the initial value of the memory 36 is O, the transform coefficient of the reference sub-block obtained by the inverse quantization unit 10 is first obtained by the addition unit 37.
Even if added with
,.

Go to the exchange section 12. At this time, the transform coefficients of the reference sub-blocks are stored in the memory 36 for each block. Regarding the error coefficient of the sub-block, the addition unit 37 uses the transform coefficient of the reference sub-block corresponding to the sub-block.
The addition with the transformation coefficient of the reference sub-block is taken by
The result is the original transform coefficient, and the data is sent to the inverse quantization section. To explain reproduction of a headline image, first, image data is reproduced using only the transform coefficients of the reference sub-block. Then,
An image of 1/4 of the original image data is reproduced. The diagram is shown in FIG. In the figure, a is the original image, and b is a quarter of the original image.

The image data reproduced in this manner can be used as a headline image in the file system search process. Also,
By decoding the error data corresponding to the transform coefficients of other sub-blocks and adding the transform coefficients of the reference sub-block that have already been reproduced, the transform coefficients of those sub-blocks are obtained, and the image is reproduced by adding them. Accordingly, the original image can also be reproduced by 11 l·-7. As described above, according to the present invention, a headline image can be efficiently created from a part of encoded data, and the original image can be reproduced by additionally reproducing the data of the headline image. In this embodiment, the number of sub-blocks constituting the block is four, but for example, there are 16 sub-blocks.
When reproducing the original image, it is possible to reproduce the image in more detail and hierarchically by first reproducing an image of 1/16th of the original image, and then reproducing an image of 1/4th of the original image.

Effects of the Invention As described above, in the present invention, it is possible to efficiently reproduce a headline image using a part of the encoded data of the original image, and to perform image reproduction in a single layer.

[Brief explanation of the drawing]

FIG. 1 is a process diagram of recording and encoding image data in an image encoding device according to an embodiment of the present invention, FIG. 2 is a block diagram of reproducing and decoding image data in the same device, and FIG. FIG. 4 is a diagram showing an example of a subsample in the device; FIG. 5 is a configuration diagram of encoded data in the device; FIG.
The figure shows 1...image memory section, 2...block readout section, 3...subblock readout section, 4...orthogonal transformation section, 5. ...Transform coefficient compression section, 6...Quantization section, 7...
Encoding unit, 8...Encoded data. Name of agent: Patent attorney Toshio Nakao and 1 other person No. 1
Figure 2 Figure 3 (b) 36 Figure 4 (b) (cn(d)
) (Fig. 5) Example of block code I L data structure (ward) ! Image coded data structure example (b) Fig. 6 (cL) (b) Fig. 7

Claims (1)

[Claims] Divide image data into blocks, further divide the block into a plurality of subblocks obtained by subsampling different pixel positions, orthogonally transform the subblocks, and encode the obtained transform coefficients. Among the sub-blocks, the transform coefficient of the first sub-block is encoded as is, and for the sub-blocks other than the first sub-block, the error between the transform coefficient and the transform coefficient of the first sub-block is encoded. An image encoding device characterized by: