JPH02150174A - Picture encoding system - Google Patents

Abstract

PURPOSE:To efficiently encode a picture signal by transforming the N components of a picture into uncorrelated components. CONSTITUTION:CMBk signals from a color scanner 1 are matrix-operated by a transformation matrix consisting of eigen vectors obtained by an eigen vector calculating part 3, and transformed into mutually uncorrelated components K1 to K4 by an uncorrelatively transforming part 4. The four pictures obtained in such a way are independently DCT-transformed by a DCT transformation part 6, and further transformation coefficients are quantized by the quantizer of a quantizing part 7. On the other hand, for the transformation matrix obtained by the eigen vector calculating part 3, the inverse matrix is calculated by an inverse matrix calculating part 5, it is delivered to a recording processing part 8 as the additional information at the time of decoding together with the quantized transformation coefficients obtained by the quantizing part 7, and it is encoded and recorded. Thus the N-color components are transformed into mutually uncorrelated components, and the printing picture can be efficiently encoded.

Description

DETAILED DESCRIPTION OF THE INVENTION (Technical field to which the invention pertains) The present invention relates to encoding for storing and transmitting images in which each pixel is represented by N color components. The present invention relates to an image encoding method that can encode an image with high efficiency even if each component is encoded independently by converting an image into N components that are unrelated to each other.

(Prior Art) Conventionally, as a method for encoding images consisting of highly correlated components, for example, for encoding images for printing, there is a method called "Compression of Images for Printing Using Adaptive Vector Quantization," Aizu, Takagi, 1986.
Annual Image Coding Symposium, cyan (C),
Magenta (M), yellow (Y).

This is a technique that utilizes the correlation between black (Bk) colors, expands a 4×4 block in the CMYBk direction, and performs vector quantization as a 16-dimensional block.

Here, vector quantization means to group multiple pixels together to form a vector, and to calculate the possible patterns of that vector.
This is done by approximating a vector with fewer patterns. With this vector quantization, an image can be approximated with the number of bits required to indicate an index indicating a smaller number of patterns.In a single image, etc., the number of patterns can be reduced by taking advantage of the characteristic that similar pixels are likely to occur in the vicinity. I'm letting you do it.

In this conventional technique, by setting vectors across CMYBk, correlations not only within one color component but also between one envelope component are utilized. In addition, since high quality is required for printed images, it is not possible to make the approximate pattern very small, and if there are many approximate patterns, a huge amount of calculation is required to select the pattern to approximate. It becomes necessary.

In addition, as in ``Study on Compression Encoding of Printing Prepress Data'' by Nakajima, Yasuin et al., 1987 Image Coding Symposium, after converting CMY from CMY to YIQ, each component is independently converted to GBTC, PO2, etc. , DCT-
There are techniques that use conventional encoding methods such as VQ.

However, the former method described above has a problem in that it takes a very long time to create a codebook because of the technique called vector quantization.

There are solutions to this problem, such as reducing the amount of codebooks depending on the required image quality, but these methods are not suitable for images that require particularly high quality, such as printed images.

Furthermore, in the latter method, since the correlation between the Bk component and the Bk component remains, there is a problem in terms of efficiency when encoding the YIQ component and Bk independently. Furthermore, regarding the YIQ component, if the three components are uncorrelated with each other, the axis that is uncorrelated in the YIQ space should match the YIQ axis, and there should be no difference between the two axes, but as shown in Figure 4. However, in reality, there is a large difference between the two axes depending on the image. That is, 3
This means that correlations between components remain, and there is a problem in encoding each component separately.

(Object of the Invention) The present invention has been made to solve the above-mentioned problems, and it is possible to encode N components for each image, which have a correlation between the components and which is problematic in terms of efficiency when encoded independently. Another object of the present invention is to encode all N components independently by converting them into uncorrelated components, thereby efficiently encoding an image signal.

(Structure of the Invention) (Characteristics of the Invention and Differences from the Prior Art) In order to achieve the above object, the present invention calculates a covariance matrix between the components of each image and projects each component in the direction of its eigenvector. The main feature is that the conversion matrix is converted into uncorrelated components and the conversion matrix is used as additional information for decoding.

The conventional technology means that according to the encoding method of the present invention, N of the image is
By converting these components into uncorrelated components, it is possible to encode each component completely independently, and since this operation is performed for each image, it is possible to The difference is that processing is possible.

(Example) An example of the present invention will be specifically described below with reference to the drawings.

It should be noted that throughout the explanation of the embodiments, parts having the same functions are given the same reference numerals, and repeated explanations thereof will be omitted.

Figure 1 shows cyan (C) and magenta (M) of the present invention.

A block diagram showing a system configuration of an encoding method when applied to a printing image consisting of yellow (Y) and black (Bk) is shown.

In Figure 1, 1 is a color scanner, 2 is CMYBk
3 is an eigenvector calculation unit for the covariance sequence, which outputs a transformation matrix composed of eigenvectors. Reference numeral 4 designates a conversion unit that makes the CMYBk components uncorrelated with each other, 5 a unit that calculates an inverse transformation matrix for the conversion matrix, 6 a DCT conversion unit, 7 a quantization unit, and 8 a recording processing unit.

In the image encoding method of this embodiment, CMYBk components of an image signal input from a color scanner 1 are sent to a covariance matrix calculation section 2. Here, first, the average value of each component is calculated, and the difference between the average value and the pixel value is calculated as C, M, Y, Bk.
A 4×4 covariance matrix is obtained by integrating between the pixels and dividing by the total number of pixels. The eigenvector of the covariance matrix is calculated by the eigenvector calculation unit 3 by diagonalizing the covariance matrix.

The CMYBk signals from the color scanner 1 are subjected to matrix calculation using a conversion matrix consisting of the eigenvectors obtained by the eigenvector calculation unit 3, and are divided into mutually uncorrelated components Kl, 2. 3. 4 in the decorrelation conversion unit 4. The four images obtained in this way are transferred to the DCT converter 6
, and the transform coefficients are quantized by a quantizer in a quantizer 7 . On the other hand, the transformation matrix obtained by the eigenvector calculation unit 3 is
The inverse matrix is calculated, and as additional information during decoding,
Together with the quantized transform coefficients obtained by the quantization unit 7,
The data is passed to the recording processing unit 8, encoded and recorded.

FIG. 2 is a block diagram showing the system configuration of another embodiment in which the present invention is applied to encoding images for printing.

In FIG. 2, 9 is an eigenvalue, an eigenvector calculation unit, 1
0 is normalization, sampling control section, 11 is normalization section, 1
2 indicates a sub-sampling section.

This embodiment is obtained by adding a normalization section 11 and a subsampling section 12 to the embodiment shown in FIG.

The four components that have become uncorrelated with each other in the decorrelation conversion section 4 are input to the normalization and subsampling control section 10, and the eigenvalue and eigenvector calculation section 9 converts the eigenvalues obtained at the same time as the eigenvectors into a certain predetermined value. By comparing,
It is determined whether to perform normalization or subsampling.

Here, normalization is for making the data fall within a certain range for convenience of operation. The components determined to be normalized are input to the normalization unit 11 and normalized by the square root of the eigenvalue. The components for which subsampling has been determined are input to the subsampling unit 12, where they are subjected to smoothing filtering to suppress aliasing noise after subsampling, and then subsampled to 1/2 in the horizontal direction. . In the case of FIG. 2, 1 is normalized, K2. 3. This shows the case where 4 is determined to be subsampling.

Furthermore, the normalized or subsampled components are
The DCT transformer 6 transforms the transform coefficients, and the quantizer 7 quantizes the transform coefficients. The quantized transformation coefficients are stored in the recording processing unit 8 along with the inverse matrix outputted from the inverse matrix calculation unit 5 and the additional information of the normalization factor outputted from the normalization unit 11.
is encoded and recorded.

(Effects of the Invention) As described above, according to the present invention, N color components are converted into mutually uncorrelated components, so each component can be encoded independently. Also, since this conversion is performed for each image, 1
This process is optimal for each image, and is suitable for encoding images for printing, where high quality is important.

As a result of simulation based on the embodiment shown in FIG. 2 of the present invention, as shown in FIG. 3, CMY-Y IQ
Compared to the method using conversion, the S/N was improved by about 1 dB.

In the above-described embodiment, the encoding method for a printing image consisting of four color components of CMYBk was described, but the present invention can also encode an image consisting of two color components, an image consisting of three color components such as RGB, and furthermore, M.Y.

Since the saturation decreases when mixing Bk colors, it can also be applied to cases where a new spot color is added to make the number of colors five or more.

In addition to the DC and T transform methods, other encoding methods such as block truncation coding (BTC), vector quantization (VQ), or Karhunen-Loewe (KL) transform can also be used. It is.

[Brief explanation of the drawing]

FIGS. 1 and 2 are block diagrams showing each system configuration when the present invention is applied to a printing image encoding method;
FIG. 3 is a diagram showing the results of simulation based on the example shown in FIG. 2, and FIG. 4 is a diagram showing the YIQ in the conventional technology.
FIG. 3 is a diagram showing the angles formed by the axis where one distribution is uncorrelated in space and the Y, I, and Q axes. DESCRIPTION OF SYMBOLS 1...Color scanner, 2...Covariance matrix calculation unit, 3...Eigenvector calculation unit, 4...Decorrelation conversion unit, 5...Inverse matrix calculation unit, 6...DCT
Conversion unit, 7... Quantization unit, 8... Recording processing unit,
9...Eigenvalue, eigenvector calculation unit, 10...
Normalization and subsampling control section, 11... normalization section, 12... subsampling section. Figure Figure Figure

Claims (1)

[Claims] When encoding an image in which each image is represented by N color components, there is a method for determining the covariance matrix between each component, and a new coordinate axis in which the direction of the eigenvector of the covariance matrix is set in the N-dimensional space. By re-expressing each component, we obtain N mutually uncorrelated components, and by using a matrix for coordinate transformation in the N-dimensional space as additional information, we obtain N mutually uncorrelated components. 1. An image encoding method comprising: means for independently encoding N components using a predetermined encoding method.