JPH08317403A - Device and method for image processing - Google Patents

Abstract

PURPOSE: To efficiently perform coding not depending on the property of an image by finding a contour line including an encoding target picture element and performing predictive coding along the contour line. CONSTITUTION: A picture element interpolator 2 assumes the presence of a picture element between the encoding target picture element and the peripheral picture element of an encoded picture element, and finds the value of the picture element, and the value of an interpolated picture element and that of the peripheral picture element of the encoding target picture element are recorded on memory 3. A maximum gradient decision device 4 selects the one whose luminance gradient is maximized in a direction advancing the outside from the encoding target picture element in radial shape. An index converter 5 generates an index in a direction intersecting orthogonally to the direction of the index representing the maximum gradient direction decided by the maximum gradient decision device 4. A predictor 6 generates a predictor by using the picture element in the direction of contour line including the encoded picture element or an interpolated picture element. A differentiator 7 generates a differential value in which the predictor is subtracted from the encoding target picture element, and a Huffman encoder 8 supplies a Huffman code to the differential value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus and method for encoding or decoding.

[0002]

2. Description of the Related Art Conventionally, DPCM (Differential Pulse Co) has been used as a reversible encoding method for multilevel images.
de Modulation) is widely known. Encoding by general DPCM includes encoded pixels around the pixel to be encoded, that is, a left pixel (referred to as a), an upper pixel (similarly b), and a left diagonal pixel (similarly c). Is used to calculate the predicted value (planar prediction) of the pixel to be encoded using, for example, the formula (a + b-c), and the difference between this and the pixel value to be encoded is encoded.

[0003]

Generally, since most of the image part is formed by a monotonous brightness gradient, a considerable part of the pixel to be coded is accurately predicted by the prediction of the pixel to be coded, Is encoded.

However, a high prediction error naturally occurs in the high frequency part in the pixel, that is, in the edge part and the thin line part. For such a part, one-dimensional prediction in the edge direction or the line extension direction is more suitable than the plane prediction.

As described above, the optimum prediction method differs depending on the properties in the image. The optimum prediction direction differs from pixel to pixel, but when such an encoding is actually attempted, the additional information indicating which prediction method is used adds encoding efficiency. I will end up.

As described above, there has been a problem that there is no conventional prediction method for efficiently encoding both the portion having a smooth gradient of pixel values and the edge portion and the thin line portion.

[0007] Further, in predictive coding by selectively using a plurality of prediction methods, additional information is required to indicate which prediction method is selected, which causes a problem of poor coding efficiency. It was

Therefore, an object of the present invention is to perform efficient coding independent of the nature of an image in order to solve the above problems.

[0009]

In order to solve the above-mentioned problems, an image processing apparatus according to claim 1 of the present invention comprises a setting means for setting a plurality of directions around a pixel to be coded, Generating means for generating reference pixel values corresponding to the plurality of directions set by the setting means, and output means for outputting gradients corresponding to the plurality of directions using the reference pixel values generated by the generating means, First selecting means for selecting a first direction based on the gradient output by the output means, and second selecting means for selecting a second direction according to the first direction selected by the first selecting means Second selecting means and the second selecting means selected by the second selecting means.
It has a coding means for coding using the reference pixel value corresponding to the direction.

[0010]

[Embodiment] In the present embodiment, attention is paid to the fact that the pixel value prediction along the contour line on the luminance curved surface is the most efficient for both the smooth curved surface in the image, the edge and the thin line, A contour line including a pixel to be encoded is found, and predictive coding is performed along the contour line.

In order to achieve this, first, the direction having the largest brightness gradient is searched for from the pixel to be coded.

Since the contour line is orthogonal to the steepest luminance gradient, the direction of the contour line can be specified by searching for it.

The reason why the contour direction is not directly searched will be described below. The brightness gradient is small in any direction close to the contour line, and the true contour line direction is hidden by the noise component on the image. On the other hand, in the method of finding the steepest gradient, the low-frequency power with a large gradient is larger than the power of noise, so it is relatively easy and accurate to know the steepest gradient direction. There is an advantage that you can.

Further, the prediction value is obtained in the contour line direction thus obtained, and the average value of a plurality of pixels in the contour line direction is used as the prediction value of the pixel to be coded, so that the variance of noise is reduced and the prediction error is reduced. The entropy of is reduced, and the coding efficiency can be improved. Examples will be described in detail below.

FIG. 1 shows an overall block diagram of a lossless encoding apparatus for multilevel images which is an embodiment of the present invention.

In FIG. 1, 1 is a line buffer,
One line containing the pixel to be encoded and four lines preceding it are stored, and this line is used when referring to the pixels surrounding the pixel to be encoded.

Reference numeral 2 denotes a pixel interpolator, which determines the value of a pixel to be coded on the assumption that there is a pixel between the pixel to be coded and the surrounding pixels of the coded pixel.

The pixel values interpolated here and the values of the pixels surrounding the pixel to be coded are recorded in the memory 3.

Reference numeral 4 denotes a maximum gradient determiner, which can select a pixel having a maximum luminance gradient in a direction (12 directions in this embodiment) radially outward from the pixel to be encoded.

Reference numeral 5 denotes an index converter, which, with respect to the index indicating the maximum gradient direction determined by the maximum gradient determining unit 4, generates an index in a direction orthogonal to this direction.

Reference numeral 6 is a predictor, which generates a predicted value by using pixels in the contour line direction including coded pixels or interpolated pixels.

Reference numeral 7 is a differencer, which generates a difference value obtained by subtracting the predicted value from the pixel value to be encoded. 8 is a Huffman encoder, which gives a Huffman code to the difference value.

Next, the entire operation will be described with reference to FIG.

Pixel data input from 101 is stored in the line buffer 1. In this embodiment, it is assumed that the scan line to be encoded and the preceding four encoded scan lines are stored.

The pixel interpolator 2 outputs from the line buffer 1
The values of the surrounding pixels of R _{0 to} R ₁₁ and A to T shown in FIG. 3 are read out. 42 pixels r ₀ (j) to r ₁₁ (j) / (j) corresponding to the 12 directions shown in FIG. 4 are _{0 to 0} using the values of these pixels.
The value of 3 or 0-2 is calculated by the following formula. That is, each value, · ri (0) = Ri ( end _{point) · r 0 (1) =} Q, r 0 (2) = S, r 0 (3) = T · r 3 (1) = D, r 3 _{(2) = K · r 6} (1) = B, r 6 (2) = F, r 6 (3) = L · r 9 (1) = H, r 9 (2) = M · r 1 (1 ) = (3I + Q) / 4, r ₁ (2) = (J +
S) / 2, r ₁ (3) = (K + 3T) /4.r ₅ (1) = (3A + B) / 4, r ₅ (2) = (E +
_{F) / 2, r 5 (} 3) = (K + 3L) / 4 · r 7 (1) = (3C + B) / 4, r 7 (2) = (F +
G) / 2, r ₇ (3) = (M + 3L) /4.r ₂ (1) = (2J + D) / 3, r ₂ (2) = (2K
+ T) /3.r ₄ (1) = (2E + D) / 3, r ₄ (2) = (2K
+ L) /3.r ₈ (1) = (2G + H) / 3, r ₈ (2) = (2M
+ L) / 3 · r ₁₀ (1) = (2N + H) / 3, r ₁₀ (2) = M · r ₁₁ (1) = P, r ₁₁ (2) = N, r ₁₁ (3) = M .

Here, in the case of the encoding target pixel position where a part of the end points cannot be referred to, as in (a) to (d) of FIG. 5, the maximum gradient determination described later only for the referenceable Ri direction. The target is the direction used in the above, and other directions are not considered.

The pixel interpolator 2 performs processing such as interpolation in accordance with the above equation, and the 42 pixel values described above are stored in the memory 3 through the bus 102.

The maximum gradient determiner 4 refers to the 42 interpolation pixel values (reference pixel values) stored in the memory 3, and then calculates the gradient by the method described later.

When the direction of Ri is "direction i", the value of the gradient in the i direction is Gi. Also, the number of interpolation pixels in the i direction is Ni,

[0030]

[Outside 1] Defines Gi. Further, in this embodiment, Hi = 1.

The maximum gradient determiner 4 finds Gi for i = 0 to 11 and outputs the value of i that gives the maximum Gi to the index converter 5.

If the input to the index converter 5 is Din and the output from the index converter is Dout, then Dou
t = (Din + 6)% 12 (% is a remainder operation). As a result, the index in the maximum gradient direction is converted into the index in the contour line direction, and the coding prediction method is determined.

The predictor 6 reads the interpolated pixel values existing in this direction from the memory 3 according to the index direction output from the index converter 5, and outputs Nindex (in
dex calculates the average value of the pixel values of the contour direction index) as a prediction value, and outputs it to the difference unit 7.

The differentiator 7 subtracts the predicted value of the pixel to be encoded from the value of the pixel to be encoded stored in the line buffer 1 and outputs this difference value to the Huffman encoder 8.

The Huffman encoder 8 performs Huffman coding on the difference value input from the differentiator 7, and outputs the coded data to the decoding side through 112.

FIG. 2 shows an example of a block diagram on the decoding side. The encoded data output from the Huffman encoder 8 is
It is decoded by the Huffman decoder 9 and output to the adder 10 as difference value data. Further, with respect to the system from the line buffer 11 to the predictor 16, the same processing as that on the encoding side is performed. Here, the line buffer 11 stores one line of decoding target pixels and four lines of decoded pixels preceding the decoding target pixel. Hereinafter, the same value as the output value of the predictor 16 to the predictor 6 is finally added to the adder 1 as the predicted value of the pixel to be decoded.
It is output to 0.

The adder 10 adds the difference value data input from the Huffman decoder to the prediction value input from the predictor 16 and reproduces and outputs it as the pixel value of the pixel to be decoded.
The reproduced and output pixel value is output to the output unit 17 through 115.
To the line buffer 1 through 116.
1 is stored as a decoded pixel value.

Since the coding prediction method in the coding method used in the above embodiments is determined by using the coded peripheral pixels without using the pixel to be coded, the coding side as well as the decoding side No additional information is required, since it can be commonly recognized.

Note that the Huffman coding of the difference value according to the embodiment may use another coding method instead.

At the position (first pixel of the line) of FIG. 5D, not the difference value but the actual pixel value is encoded and sent, and the decoding side decodes it as the actual pixel value. It is also possible to control.

For the first four lines of the image,
By setting an index determined by the predictor 6, it is possible to control so as to send a difference value with respect to a preset direction.

Further, in the above-described embodiment, the maximum gradient direction is detected and the predicted value is determined by using the pixels in the direction orthogonal to this direction. However, the direction of the center of the direction having a gradient of a predetermined value or more is detected. The method can be selected as appropriate.

[0043]

According to the present invention, it is possible to perform efficient coding independent of the nature of an image.

[Brief description of drawings]

FIG. 1 is an overall block diagram of an image encoding device according to an embodiment.

FIG. 2 is an overall block diagram of an image decoding apparatus according to an embodiment.

FIG. 3 is a diagram illustrating surrounding pixels referred to in image prediction according to the embodiment.

FIG. 4 is a diagram for explaining the operation of the pixel interpolator 2 of FIG.

FIG. 5 is a diagram illustrating a case where there are surrounding pixels that cannot be referenced.

[Explanation of symbols]

 1 line buffer 2 pixel interpolator 3 memory 4 maximum gradient determiner 5 index converter 6 predictor 7 differentiator 8 encoder

Claims (16)

[Claims] 1. A setting means for setting a plurality of directions centering on a pixel to be encoded, a generating means for generating reference pixel values corresponding to the plurality of directions set by the setting means, and a generating means for generating the reference pixel values. Output means for outputting gradients corresponding to the plurality of directions by using the reference pixel values thus obtained; first selecting means for selecting a first direction based on the gradients output by the output means; A second selecting means for selecting the second direction according to the first direction selected by the first selecting means; and a reference pixel value corresponding to the second direction selected by the second selecting means. An image processing apparatus having an encoding unit for performing encoding using the image processing apparatus. 2. The image processing apparatus according to claim 1, wherein the pixel to be encoded is not used for outputting the gradient by the output unit. 3. The image processing apparatus according to claim 1, wherein the reference pixel value is a brightness value. 4. The image processing apparatus according to claim 1, wherein the first direction and the second direction are orthogonal to each other. 5. The image processing apparatus according to claim 1, wherein the reference pixel value includes a fictitious pixel value. 6. The coding apparatus according to claim 1, wherein Huffman coding is used for coding by the coding means. 7. The image processing apparatus according to claim 1, wherein the encoding by the encoding unit is performed for each pixel. 8. A plurality of directions are set around a pixel to be encoded, reference pixel values corresponding to the plurality of set directions are generated, and the generated reference pixel values are used to correspond to the plurality of directions. Output a gradient, select a first direction based on the output gradient, select a second direction according to the selected first direction, and a reference pixel corresponding to the selected second direction An image processing method characterized by performing encoding using a value. 9. A calculation means for calculating a gradient of a pixel value in a plurality of directions centering on a pixel to be encoded, and a gradient having a desired gradient among the gradients calculated by the calculation means is orthogonal to the direction having the desired gradient. An image processing apparatus comprising: an encoding unit that performs predictive encoding based on a direction. 10. The image processing apparatus according to claim 9, wherein the desired gradient has a maximum value of the gradient calculated by the calculating means. 11. The image processing apparatus according to claim 9, wherein the calculation means uses the pixel values around the pixel to be encoded, without using the pixel value to be encoded. 12. A gradient of pixel values is calculated for a plurality of directions centering on a pixel to be encoded, and predictive coding is performed based on a direction orthogonal to a direction having a desired gradient among the calculated gradients. Image processing method for performing. 13. Setting means for setting a plurality of directions centering on a pixel to be decoded, generating means for generating reference pixel values corresponding to the plurality of directions set by the setting means, and generating by the generating means. Output means for outputting gradients corresponding to the plurality of directions by using the reference pixel values thus obtained; first selecting means for selecting a first direction based on the gradients output by the output means; A second selecting means for selecting the second direction according to the first direction selected by the first selecting means; and a reference pixel value corresponding to the second direction selected by the second selecting means. An image processing apparatus having a decoding means for performing decoding using the image processing apparatus. 14. A plurality of directions are set around a pixel to be decoded, reference pixel values corresponding to the set plurality of directions are generated, and the generated reference pixel values are used to correspond to the plurality of directions. Output a gradient, select a first direction based on the output gradient, select a second direction according to the selected first direction, and a reference pixel corresponding to the selected second direction An image processing method characterized by performing decoding using a value. 15. A calculating means for calculating a gradient of a pixel value in a plurality of directions centering on a pixel to be decoded, and a gradient having a desired gradient, out of the gradients calculated by the calculating means, is orthogonal. An image processing apparatus comprising: a decoding unit that performs decoding based on a pixel value in a direction. 16. A gradient of pixel values is calculated for a plurality of directions centered on a pixel to be decoded, and based on pixel values in a direction orthogonal to a direction having a desired gradient among the calculated gradients. Image processing method for decoding.