JPH04326667A - Compression and expansion device for picture data - Google Patents

Abstract

PURPOSE:To process a conversion coefficient under the same condition independently of a bit number change in a data of a digital picture. CONSTITUTION:A digitized picture data is subject to orthogonal transformation by an orthogonal transformation/orthogonal inverse transformation arithmetic section (DCT/IDCT) 10 and a conversion coefficient obtained by the orthogonal transformation is quantized, coded and compressed. In this system, bit shift in which a bit number (b) of a picture data is made coincident with a maximum bit number (p) able to be operated in the arithmetic section 10 at a position A or B before the arithmetic section 10 or a bit number (r) of the conversion coefficient is made coincident with a bit number (q) of the conversion coefficient obtained when a picture data of the bit number (p) is subject to orthogonal transformation at a position C after the arithmetic section 10 is implemented. Thus, the bit number of the conversion coefficient to be quantized and coded is unified.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data compression/expansion apparatus, and more particularly to an apparatus configured to obtain compressed data by quantizing and encoding transform coefficients obtained by orthogonally transforming image data.

0002

[Background Art] With the advancement of digital technology, it has become common practice in recent years to digitize gradation images such as X-ray images, store them, transmit them electronically, and perform various digital image processing. It has become. However, since a gradation image has a larger amount of information than a binary image, there is a problem in that the large amount of data when digitizing a gradation image causes an increase in transmission time and an increase in memory capacity. In particular, for medical X-ray images, the number of pixels and the number of bits required for each pixel when digitized is 4, for example, for a chest X-ray photograph.
It is huge, with 100,000 to 16 million pixels and 8 to 12 bits, and is inefficient in storing and transmitting data.

[0003] Therefore, image compression technology that compresses a huge amount of data of gradation images including medical images is attracting attention. Such image compression techniques can be broadly classified into reversible compression and irreversible compression, but reversible compression is 1/2 to 1/3
Since only a relatively low compression rate can be expected, irreversible compression methods that can achieve a high compression rate of 1/5 or higher, particularly orthogonal transform coding, are attracting attention.

Orthogonal transform encoding involves dividing the entire image into small blocks, applying orthogonal transform such as discrete cosine transform (DCT) to each of these blocks, and quantizing and encoding the resulting transform coefficients. This is one of the irreversible compression methods and is known as the most suitable compression method for compressing gradation images. When decompressing compressed data compressed using orthogonal transform as described above to restore image data, the compressed data is decoded and dequantized to obtain transform coefficients, and then the orthogonal inverse By performing the conversion, restored image data can be obtained.

[0005]

[Problem to be solved by the invention] By the way, in medical gradation images, the dynamic range of image data (the number of bits representing the gradation of each pixel) differs depending on the imaging method such as X-ray images, CT images, MRI images, etc. Furthermore, even the same X-ray image may have three dynamic ranges: 8 bits/pixel, 10 bits/pixel, and 12 bits/pixel, depending on the device that took the image, for example.

[0006] Here, as mentioned above, image data compression and
It is preferable that the decompression device be capable of compressing and decompressing image data having different dynamic ranges as described above. However, when performing image compression using orthogonal transformation such as discrete cosine transformation, if image data with different dynamic ranges is directly orthogonally transformed, the dynamic range of the resulting transform coefficients depends on the dynamic range of the image data. This is because the processing conditions before encoding the transform coefficients, such as transform coefficient selection parameters and quantization parameters, must be set according to the dynamic range (number of bits) of the transform coefficients. When the dynamic range of the conversion coefficient changes as described above, it becomes necessary to adjust the parameters each time, and there is a problem in that it is not possible to respond flexibly to changes in the dynamic range of image data.

The present invention has been made in view of the above-mentioned problems, and is capable of encoding transform coefficients even if the dynamic range of image data changes in a compression/expansion device that compresses and expands data using orthogonal transform. It is an object of the present invention to provide an image data compression/expansion device that does not require adjustment of processing conditions up to the present time and can maintain high calculation accuracy in orthogonal changes.

[0008]

[Means for Solving the Problems] Therefore, in the present invention, transform coefficients obtained by performing orthogonal transformation on digitized image data are quantized and encoded to obtain compressed data, and the compressed data is decoded and inverted. In an image data compression/decompression device configured to perform quantization, perform orthogonal inverse transform, and decompress image data, orthogonal transform is performed so that the number of bits of the transform coefficient obtained by performing orthogonal transform matches a predetermined number of bits. Performs a bit shift operation on the previous image data or the transform coefficient obtained by orthogonal transformation, then performs compression processing, and when decompressing, the bit shift operation in the opposite direction to the compression process returns to the number of bits of the original image data, and decoding - It is configured to perform on the transform coefficients obtained by inverse quantization or the image data restored by orthogonal inverse transform.

[0009] Here, it is preferable that the predetermined number of bits is the number of bits of a transform coefficient obtained by performing orthogonal transform on image data having the maximum number of bits that can be computed without causing an overflow in orthogonal transform computation.

0010

[Operation] According to such an image data compression/expansion device, when compressing an image, the number of bits of the image data before orthogonal transformation is made to match the predetermined number of bits, or the number of bits of the transformation coefficient obtained by orthogonal transformation is A bit shift operation is performed to match a predetermined number of bits. Therefore, even if the number of bits of image data changes, the number of bits of transform coefficients obtained by orthogonal transformation will be unified to a predetermined number of bits. For this reason, parameters related to compression processing after orthogonal transformation, such as coefficient selection parameters and quantization parameters of transform coefficients, are
There is no need to make adjustments in response to changes in the number of bits in the conversion coefficient.

Furthermore, during decompression, the number of bits of the original image data before compression can be restored by performing a bit shift operation in the opposite direction to that during compression. Here, the predetermined number of bits is the number of bits of a transform coefficient obtained by performing orthogonal transformation on image data with the maximum number of bits that can be calculated without causing overflow in orthogonal transform operation, and By adjusting the number of bits to match the maximum number of bits, it is possible to maintain high calculation accuracy in orthogonal transformation calculations.

0012

[Examples] Examples of the present invention will be described below. FIG. 1 is an overall configuration diagram of a system equipped with an image data compression/expansion device according to the present invention. This system is composed of a host computer 1 and a data compression/expansion device 2. The digitized image data or compressed encoded data (compressed data) is transferred to the data compression/expansion device 2, and the compressed encoded data or restored image data is transferred from the data compression/expansion device 2 to the host computer 1 side. It will be sent back to.

Here, the host computer 1 and the data compression/expansion device 2 are connected via a bus 3, and the bus 3 is a bidirectional data bus for transferring data such as image data and command data. It has a control bus for sending and receiving control signals associated with data transfer. The data compression/decompression device 2 includes an image buffer 4 that temporarily stores digital image data to be compressed or restored image data (expanded image data), and an image buffer 4 that temporarily stores compressed coded data or coded data that is the source of restoration. It is equipped with a code buffer 5 for storing data in the code buffer 4,
5 is connected to the bus 3 via a bus controller 6, and data buffers 4, 5 are connected to the bus 3 via a bus controller 6.
5 or data transfer from the buffers 4 and 5 to the host computer 1 side.

The data compression/expansion engine 7 is a part that performs irreversible compression/expansion of image data using orthogonal transformation such as discrete cosine transformation (DCT).
Image compression is performed using the following steps. First, image data is divided into block images consisting of n pixels x n pixels, and then a two-dimensional discrete cosine transform (DCT) is applied to each block image, and a transform coefficient matrix F (u, v) (u ,v=0,1,2,...n-1
).

[0015] Since the transform coefficients are real numbers, such transform coefficients cannot compress data as they are, so the number of levels required to represent the coefficients is reduced by rounding (quantizing) the coefficient values. Furthermore, the compression ratio is increased by truncating conversion coefficients with small amplitudes. Next, encoding is performed in which code words with different code lengths (variable code length) are assigned to the quantized coefficients.If a shorter code is assigned to a coefficient with a higher probability of occurrence, then
The quantized coefficients can be compressed into a small amount of data, and the image data is compressed by completing the encoding process.

[0016] In order to reproduce a restored image from compressed data, it is sufficient to decipher (decode) the encoded data, replace it with quantized data (inverse quantization), and perform decompression processing that performs orthogonal inverse transformation on this data. . The operations of the buffers 4 and 5, the bus controller 6, and the data compression/expansion engine 7 in the data compression/expansion device 2 are controlled by a CPU 8 built into the data compression/expansion device 7.

By the way, in this embodiment, taking the image data compression as an example, A (before the image buffer circuit 4) and B (after the image buffer circuit 4) shown in FIG.
A dynamic range conversion circuit 9 is inserted at either a location (before the orthogonal inverse transform calculation unit 10) or C (after the orthogonal transform/orthogonal inverse transform calculation unit 10 and before the quantization unit). As shown in FIG. 3, the dynamic range conversion circuit 9 inputs actual image data or the number of bits of conversion coefficients, or control data for the number and direction of shifts, and converts the data input via the data bus. This circuit performs a bit shift operation based on the control data and outputs data after the bit shift operation.

Here, when the actual number of bits of image data or conversion coefficients is input to the dynamic range conversion circuit 9 as control information, the number of bits of image data or conversion coefficients set in advance and the actual number of bits of image data or conversion coefficients are The shift direction and number of shifts are determined by comparing the number of bits of 9 may be input.

In FIG. 2, the dynamic range conversion circuit 9 is inserted at position A or B, that is, before the conversion in the orthogonal transform/inverse orthogonal transform (DCT/IDCT) calculation unit 10 in the data compression/expansion engine 7. In this step, a bit shift operation is performed so that the number of bits of the image data to be orthogonally transformed matches a predetermined number of bits. Here, the number of bits of the original image data is b, orthogonal transformation/
The maximum number of bits of image data that can be calculated without causing an overflow in the orthogonal inverse transform calculation section 10 is p, and the image data with the maximum number of bits p is orthogonally transformed/orthogonal inverse transform calculation section 1.
The number of bits of the transform coefficient obtained by performing orthogonal transformation with 0 is q, and the number of bits of the transform coefficient obtained by orthogonally transforming the image data with the number of bits b in the orthogonal transform/orthogonal inverse transform calculation unit 10 is r. Then, when the dynamic range conversion circuit 9 is inserted at position A or B, the conversion circuit 9 performs the following bit shift operation.

・If b<p, move b to the MSB side (p-b
) bit shift (see Figures 4 and 7). - If b=p, do not bit shift b. - If b>p, shift b bits to the LSB side (b-p) (see FIGS. 5 and 6).

That is, when the dynamic range conversion circuit 9 is inserted at the position A or B, the number of bits of the image data input to the orthogonal transformation/orthogonal inverse transformation calculation section 10 is changed by orthogonal transformation/orthogonal inverse transformation. The bits are shifted so as to match the maximum number of bits p of image data that can be calculated without causing an overflow in the calculation unit 10. Therefore, even if an image with a bit number exceeding the maximum bit number p is input as image data to be compressed, data overflow will not occur in the orthogonal transform/orthogonal inverse transform calculation unit 10, and the transform coefficients Since the number of bits is constant at q regardless of the number of bits of the input image data,
There is no need to change the processing conditions such as quantization before encoding the transform coefficients depending on the number of bits of the image data (transform coefficients), and it is possible to always perform processing such as quantization under constant processing conditions. can.

The number of bits of the transform coefficient obtained by the orthogonal transform performed by the orthogonal transform/orthogonal inverse transform calculation unit 10 changes depending on the number of bits of the input image data, and when the number of bits of the transform coefficient changes, the sign changes. Although it is necessary to adjust the processing conditions such as quantization up to quantization depending on the change in the number of bits of the transform coefficient, the image input to the orthogonal transform/orthogonal inverse transform calculation unit 10 by the bit shift operation as described above. By unifying the number of bits of data, the number of bits of conversion coefficients is unified, so even if the number of bits of input image data varies, such as 8 bits/pixel, 10 bits/pixel, or 12 bits/pixel. There is no need to change processing conditions such as quantization, and image data with different bit numbers can be processed as is.

In addition, the number of bits of the image data is maintained at the maximum number of bits that can be processed by the orthogonal transform/orthogonal inverse transform calculating section 10.
Since the orthogonal transform operation is performed using the largest number of bits, it also has the effect of maintaining high calculation accuracy in the orthogonal transform. 4 to 7, the bit shift operation of image data is performed when the number of bits b of the input image data is larger than the maximum number of bits p that can be processed in the orthogonal transform/orthogonal inverse transform calculation unit 10 (b> p) and if small (b<
Furthermore, the actual state of bit shifting is shown depending on whether the image data is represented as SIGNED data or UNSIGNED data.

For example, FIG. 4 shows a case where the image data is expressed as SIGNED data with the MSB as the sign bit (in the case of this embodiment, the case of expression of sign bit (MSB) + data absolute value will be explained). Yes, and p
>b. Here, if p-b=3, the image data is shifted 3 bits to the MSB side during compression, and the LSB side of the original image data is shifted by 3 bits to the MSB side. If 3 bits that are 0 are added to , the number of bits of the image data processed by the orthogonal transform/orthogonal inverse transform calculation unit 10 will apparently be p, and the number of bits of the transform coefficient obtained after the transform will be q. If the number of bits of image data is increased during compression in this way and then compression processing is performed, the LSB side should be deleted so that the 3 bits on the LSB side of the image data restored by orthogonal inverse transformation are deleted during decompression. By shifting the image data by 3 bits, the number of bits of the original image data can be restored.

On the other hand, when the dynamic range conversion circuit 9 is inserted at the position C in FIG. The following bit shift operation is performed so that the number r of bits of the transform coefficient obtained by orthogonal transformation matches the predetermined number q of bits before quantization.

・If b<p, move r to the MSB side (q−r
) bit shift (see Figure 8). - If b=p, do not bit shift r. When the dynamic range conversion circuit 9 is inserted at the position A or B, the number of bits of the image data before conversion is unified, so that the number of bits of the conversion coefficient after conversion is unified to q. However, when inserting the dynamic range conversion circuit 9 at position C, the number of bits r of the actually obtained conversion coefficient (which changes according to the change in the number of bits of image data) is changed to a predetermined number of bits q. By unifying the number of bits of the transform coefficients, the processing conditions such as quantization until the transform coefficients are encoded can be adjusted to match the image data. This has the effect that it is no longer necessary to change the bit number in response to a change in the number of bits.

However, if the number of bits b of the image data input to the orthogonal transform/orthogonal inverse transform calculation unit 10 exceeds the maximum bit number p (b>p), the orthogonal transform/orthogonal inverse transform calculation unit 10 Since data overflow will occur during orthogonal transformation, the dynamic range conversion circuit 9 is placed at position A or B rather than position C in FIG.
It is desirable to insert

As described above, when the dynamic range conversion circuit 9 performs a bit shift operation to unify the number of bits of image data or conversion coefficients during compression,
When decompressing, it is necessary to return the number of bits to the number of bits of the original image data, and it is necessary to use transform coefficients obtained by decoding encoded data (compressed data) and inverse quantization, or by orthogonal inverse transformation of the transform coefficients. Either one of the image data may be bit-shifted by the same number of bits in the opposite direction to the shift direction during compression (see FIGS. 4 to 8).

Furthermore, although the SIGNED data has been explained using the absolute value representation of the sign bit + data, it is also possible to use one's complement representation or two's complement representation, and in that case, the same applies to the above embodiments. Similar effects can be obtained by performing similar treatments.

[0030]

[Effects of the Invention] As explained above, according to the image data compression/expansion apparatus according to the present invention, it is possible to unify the number of bits of transform coefficients after orthogonal transformation without being affected by changes in the number of bits of input image data. This eliminates the need to change processing conditions such as quantization until the transform coefficients are encoded depending on changes in the number of bits in the image data, and has the effect of allowing image data with different numbers of bits to be processed as is. .

[0031] Furthermore, if a bit shift is performed to match the number of bits of the actual image data to the maximum number of bits that can be calculated without overflowing in the orthogonal transform operation, the number of bits of the image data subjected to the orthogonal transform operation can be maximized. This has the effect that the calculation accuracy of orthogonal transformation can be maintained at a high level.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the overall configuration of a system including an image data compression/decompression device according to the present invention.

FIG. 2 is a block diagram illustrating locations where bit shift operations are performed in the process of compressing image data.

FIG. 3 is a block diagram showing input data of a dynamic range conversion circuit that performs a bit shift operation.

FIG. 4 is a digital data state diagram showing a bit shift operation of image data (SIGNED data; sign bit + data absolute value).

FIG. 5 is a digital data state diagram showing a bit shift operation of image data (SIGNED data; sign bit + data absolute value).

FIG. 6 is a digital data state diagram showing a bit shift operation of image data (UNSIGNED data).

FIG. 7 is a digital data state diagram showing a bit shift operation of image data (UNSIGNED data).

FIG. 8 is a digital data state diagram showing a bit shift operation of conversion coefficient data (SIGNED data; sign bit + data absolute value).

[Explanation of symbols]

1 Host computer 2 Data compression/expansion device 3 Bus 4 Image buffer circuit 5 Code buffer circuit 6 Bus controller 7 Data compression/decompression engine 8 CPU

Claims (2)

[Claims] Claim 1: A transform coefficient obtained by performing orthogonal transformation on digitized image data is quantized and encoded to obtain compressed data, and the compressed data is decoded and inversely quantized and orthogonally inversely transformed to obtain an image. In an image data compression/decompression device configured to decompress data, image data before orthogonal transformation or after orthogonal transformation is applied so that the number of bits of a transform coefficient obtained by performing orthogonal transformation matches a predetermined number of bits. After performing a bit shift operation on the obtained transform coefficient, compression processing is performed, and during decompression, the bit shift operation in the opposite direction to that during compression returns to the number of bits of the original image data, and the conversion obtained by decoding and dequantization. 1. An image data compression/expansion apparatus, characterized in that the image data compression/expansion apparatus is configured to perform compression/expansion on image data restored by coefficients or orthogonal inverse transformation. 2. A claim characterized in that the predetermined number of bits is the number of bits of a transform coefficient obtained by orthogonally transforming image data with the maximum number of bits that can be calculated without causing an overflow in an orthogonal transform operation. Item 1. The image data compression/expansion device according to item 1.