JPH0275275A - Block coding device for multilevel picture - Google Patents

Abstract

PURPOSE:To save coding quantity by approximating a block including a contour line or the like with a few representative levels even when a picture element level in a block is distributed in a wide level range. CONSTITUTION:A maximum picture element level and a minimum picture element level in a fetched block 502 are obtained by a gradation change detection circuit 202, which obtains a difference between the maximum value and the minimum value and each picture element level of the fetched block 502 is given to a quantizing circuit 204 together with the difference. Then a mode identification circuit 208 decides modes in which a mode having a singular picture element level representing the fetched block 502, a mode B represented by the two levels and a mode C represented by four levels. Even when a picture element level in the fetched block 502 is distributed in a wide level range, when a contour line is included, the two representative levels are decided. Thus, the coding quantity of resolution components phi1, phi2 is saved.

Description

[Detailed Description of the Invention] [Table of Contents] Overview Industrial Field of Application Conventional Technology Problems to be Solved by the Invention Means for Solving the Problems Actions Examples Effects of the Invention [(Already needed) Block codes for multivalued images The purpose of the present invention is to provide a block encoding device that can efficiently compress multivalued images including contour lines, a means for capturing images with multilevel pixels in block units, and a means for capturing images with multilevel pixels in units of blocks. A means for instructing the generation of a pixel level that is the only representative of the block when the pixel levels are distributed in a narrow range, and a means for instructing the generation of a block representative level when the pixel levels of the captured block are distributed over a wide range and there are few intermediate level pixels. A means for instructing generation of a small number of block representative levels, a means for instructing generation of a large number of block representative levels when the pixel levels of the captured block are distributed over a wide range (from 1 to 2), and there are many intermediate level pixels, and a means for instructing generation of a large number of block representative levels. means for generating data based on each pixel level of the captured block, means for quantizing each pixel level of the captured block to one of the block representative levels, means for determining the distribution width of the block representative level, and a standard method for the captured block. There are means for determining the pixel level, and means for encoding the quantization content, the distribution width of the block representative level, and the standard pixel level at the pixel level.

[Industrial Field of Application] The present invention relates to an apparatus for block encoding multivalued images.

A multivalued image input from a color camera or scanner is formed by multilevel pixels, and therefore the amount of data in the multivalued image is enormous compared to numbers and characters.

Therefore, in order to transfer a multilevel image or process it on a computer, it is necessary to encode it efficiently.

Therefore, multivalued images are encoded using this type of apparatus in an effort to make them more compact.

[Prior Art] Regarding this type of device, proposals have been made in Japanese Patent Application No. 142501/1983, and the procedure of conventional block encoding processing is explained in Fig. 5 (Research by the Institute of Image Electronics Engineers). 86-02-05 (Refer to ``Multi-gradation Block Coding of Grayscale Images'').

The multivalued image 500 in the figure is divided into a large number of blocks 502, which are sequentially imported into the apparatus.

The wrap lock 502 is a rectangular area made up of NXN (N=4) (that is, 16) pixels, and each pixel is expressed in one of 256 levels.

Then, when any block 502 is loaded into the device, the encoding parameter TIT T2 (TI<T2) is selected.

Further, the maximum and minimum pixel levels MAJL and MIN-L are searched from the block 502 taken into the apparatus, and the difference therebetween is determined.

Next, the difference between these maximum pixel levels MA, XL and the minimum pixel level MIN-L is compared with the encoding parameters TI, T2, and if MAX-L-MIN L≦T1, mode A is selected. selected, all pixel levels within the block are quantized to a single level P9.

Further, in the case of TI<MAX-L M I N -L≦T2, mode B is selected and the pixel level within the block is quantized to one of two levels P++ P2.

If T2<MAX-L-M I N -L, mode C is selected, and each pixel level in the block is equal to four levels QlI Q2. Q3. Q4 is quantized.

The multi-value used when performing this pixel level quantization can be understood by the following equation.

<Mode A> P o-AVE (X 1j) = La (φ +: l j-0+ (φ2) 1 no O (for all i, j) Damn B> P+-AVE (Xij≧ (MAX-L+M I N
-L)/2)Pa-AVE (Xij<(MAX-L+
M I N-L)/2) L a-(P++P2)/2 Ld”P+-Pa (φ1) 1j-0 (When Xij≧(MAX-L+M I N-L)/2) (φ1) 1": 1 (If Xij<(MAX, -L+M I N-L)/2) (φ2) 1"・0 (For all j, j) Damn C> Q+=AVE (Xij≧ (3MAX-L+M r N
-L)/4)O4-AVE (Xij<(MAX-L+
3M I N-L)/4) La=(QI+QA)/2 Ld22 (Q I-O4)/3 (φ+)z=o, (φ2)++”0(Xij≧La
+Ld/2) (φI)1j:O2(φ2)ll”1 (If L a+Ld/2>Xij≧La) (φ1)I
J=1. (φ2) lI”0 (when L a>Xij>=L a−L d/2) (φ
t)++-L (φ2)(j-1(La-Ld/2>
Xij) In this way, the maximum pixel level M in the acquisition block 502
The number of levels of pixel level quantization is single depending on the difference between AX-L and the minimum pixel level MIN-L.

Specified as either a small number (=2) or a large number (=4).

The reference level La is then compared with the previous one, and the difference is subjected to variable length encoding processing.

Further, variable length encoding processing is performed on the level interval Ld after nonlinear quantization.

Further, a pair of bit planes representing the above-mentioned quantization contents and forming resolution components φ1 and φ2 are encoded by the MMR method, and the encoded signal and the reference level La are encoded by the MMR method.

Code signals with a level interval Ld are combined for transfer or the like.

As described above, conventionally, the number of levels representing this block 502 is determined only from the difference between the maximum pixel level MAX-L and the minimum pixel level MIN-L of the acquisition block 502, and the block Each pixel level within 502 is quantized.

Therefore, when the amplitude of the pixel level within a block is small, all pixel levels are approximated by Lebel, when the amplitude is large, it is approximated by a large number of levels, and when the amplitude is intermediate, it is approximated by a small number of levels.

[Problem to be solved by the invention] In FIG. 6, the level (gradation) in each block 502
The mountain on the right side of the figure corresponds to a pattern in which brightness and hue gradually change, and the gradation gradually changes.

In this part, the difference between the maximum pixel level MAX-L and the minimum pixel level MIN-L is large, and therefore each pixel level is quantized into a large number of levels.

Also, the mountain part on the left side in the same figure corresponds to the contour line, etc.
The gradation changes rapidly.

Even in the steep mountain portion, the difference between the maximum pixel level MAX-L and the minimum pixel level MIN-L is large, so each pixel level is quantized into a large number of levels.

Here, since the mountain portion on the left side of the figure corresponds to the contour line, this portion can be sufficiently restored even if they are quantized at a small number of levels, and there is no need to approximate -10- at a large number of levels.

Therefore, in the past, the amount of code for resolution components corresponding to contour lines and the like increased, although the image quality during restoration did not improve.

The present invention has been made in view of the above-mentioned conventional problems,
The purpose is to provide an apparatus that can efficiently compress multivalued images containing contour lines and the like.

[Means for Solving the Problems] In order to achieve the above object, the device according to the present invention has the following features:
It is configured as shown in the figure.

The means 10 in the figure captures an image formed by pixels of multi-level expression in blocks, and when the pixel levels of the captured blocks are distributed within a narrow range,
The means 2 instructs generation of a pixel level that is the only representative of the block.

Furthermore, the pixel level of the captured block is distributed over a wide range,
When there are few intermediate level pixels, the means 14 instructs generation of a small number of representative pixel levels of the block.

Further, when the pixel levels of the captured block are distributed over a wide range and there are many intermediate level pixels, the means 16 instructs generation of a large number of representative pixel levels of the block.

Means 18 then generates the number of block representative levels indicated by any of means 12, 14, 16 based on each pixel level of the captured block.

The means 20 also quantize each pixel level of the captured block to one of the block representative levels generated by the means 18.

Furthermore, the means 22 determines the distribution width of the block representative level, and the means 24 determines the standard pixel level of the captured block.

Then, in the means 26, the pixel level quantization content by the means 20, the distribution width of the block representative level by the means 22, and the standard pixel level by the means 24 are encoded.

[Operation] In the present invention, even if the pixel levels of the captured block are widely distributed, when the number of intermediate level pixels is small, a large number of representative pixel levels of the block are generated. However, a small number of them are generated.

[Embodiments] Hereinafter, preferred embodiments of the apparatus according to the present invention will be described based on the drawings.

FIG. 2 shows the block configuration of the embodiment, and data of an image 500 obtained by a camera, scanner, etc. and expressed by pixels with multiple levels (for example, 256 gradations) is written into a buffer memory 200. It will be done.

Then, the image 500 in the buffer memory 200 is captured in block units (4×4 pixels) by the gradation change detection circuit 202, and the maximum pixel level M in the capture block 502 is
AX-L and the minimum pixel level MIN-L are determined by the gradation change detection circuit 202.

Furthermore, this gradation change detection circuit 202 detects the maximum value MAX
The difference between -L and the minimum value MIN-L is calculated, and together with the difference, each pixel level of the acquisition block 502 is determined by the quantization circuit 2.
Given on 04.

The quantization circuit 204 uses a value T that divides the maximum pixel level MAX-L and the minimum pixel level MII'lL into three regions.
IQ+T 20 are generated and their values T+o+T
2Q and each pixel level of acquisition block 502 are compared.

This determines which of the three regions delimited by the value T IO+T 20 each pixel level of the acquisition block 502 belongs to, and as a result, each pixel level is quantized to one of the three levels.

In this way, each pixel level of the acquisition block 502 is set to 3.
The state detection circuit 206
So for each region (i.e. for each quantization level),
The number of pixels within the area is determined.

Then, in the mode identification circuit 208, mode A (one gradation approximation) has a single pixel level representing the capture block 502, and mode B (two gradation approximation) has two levels.
, or mode C (four-gradation approximation), which is represented by four levels, is determined.

The determination conditions are: Mode A MAX-L-M I N-L <Ta Mode B T8≦MAX, -L-MIN-L Katn Il<
N th mode C T8≦MAXj--M I N, -L Ka nII
Indicated by ≧Nth.

The value T6 is a threshold value for determining the maximum amount of change in pixel level within the acquisition block 502, and the value Nth is the threshold value for determining the maximum amount of change in pixel level within the acquisition block 502.
Threshold block 502 with MOrR for determining the number of pixels belonging to the intermediate region among the pixels (quantized into levels)? Approximately 10% of the total number of pixels = 16)
It is said that

That is, in the mode identification circuit 208, the difference between the maximum pixel level MAJL and the minimum pixel level MIN-L is the value T.
8 and the pixel levels of acquisition block 502 are distributed in a narrow range, mode A is selected.

Also, the maximum pixel level MAX)L and the minimum pixel level MIN
-L is greater than or equal to the value 18, the number n11 of pixels belonging to the intermediate level range is compared with the value 1'lth, and the number n0 of pixels in the intermediate level is greater than the number indicated by the value Nth. Mode B is selected when the noise is small, and mode C is selected when the noise is large.

In this way, the mode identification circuit 208 selects mode A.
When B or C is selected, this selection mode A,
The number (1, 2, or 4) of block representative levels indicated by B or C is determined by the representative gradation determination circuit 210.

The representative level of the capture block 502 determined by the representative gradation determination circuit 210 is provided to the comparison circuit 214 via the memory circuit 212 and compared with the level of each pixel within the capture block 502.

As a result, these pixel levels are quantized to one of the block representative levels, which results in the aforementioned resolution components φ1. φ2 is obtained.

And resolution component φ1. φ2 is the buffer memory 216.
218 to the encoding circuit 220 and encoded.

Furthermore, the resolution component φ1. The encoded data of φ2 is stored in memory 2.
22 and 224 and is supplied to a signal combiner 226.

Also, mode A is set in the representative gradation determining circuit 210.

When the number of block representative levels corresponding to B or C is obtained, the difference value generating circuit 228 calculates the level interval Ld that is the distribution width of the block representative levels, and the level interval Ld
is provided to the difference value encoding circuit 230.

In the difference value encoding circuit 230, the level interval Ld is encoded in a conventional manner, and the encoded data is supplied to the signal synthesizer 226 via the buffer memory 232.

Then, the reference level generation circuit 234 obtains a reference level La, which is a standard pixel level of the capture block, and the reference level La is encoded by the reference level encoding circuit 23G in the same manner as in the prior art.

Further, the encoded data is provided to a buffer memory 238 and supplied to a signal combiner 226.

This signal combiner 226 combines the encoded data provided from the memories 222, 224 and the buffer memories 232, 238, and the combined signal is transferred or recorded as is.

This embodiment has the above configuration, and its operation will be explained below.

When each pixel level of the block 502 taken from the buffer memory 200 to the gradation change detection circuit 202 is almost constant, the maximum pixel level MAX-
The difference between L and the minimum pixel level MIN-ri is less than the value T8.

Therefore, mode identification circuit 208 selects mode A,
A single block representative level is determined by the representative gradation determining circuit 210.

That is, if each pixel data in the block 502 taken into the gradation change detection circuit 202 is distributed in a narrow level range, they are quantized to one level,
It is encoded by the encoding circuit 220.

In addition, block 5 taken into the gradation change detection circuit 202
Maximum pixel level MAX-L and minimum pixel level MI of 02
If the difference from NL is greater than or equal to the value Ta and the pixel levels are distributed over a wide range of levels, mode B (two-tone approximation) or mode C (four-tone approximation) is used for mode identification. Selected in circuit 208.

In FIG. 3, the quantization operation of the pixel level performed in the quantization circuit 204 prior to the mode selection by the mote identification circuit 208 is explained. There is.

As mentioned above, this pattern has outlines etc. in blocks 50.
In this case, the pixel levels are concentrated at the minimum pixel level MIN-L and the maximum pixel level MAX-L, as shown in FIG. 4(A).

Therefore, there are almost no pixel levels at the intermediate level between these pixel levels MIN-L and MAX-L. Therefore, when the pixel levels are divided into three level ranges as shown in Figure 3, the pixel levels are grayed out in the middle level range. The number of pixels becomes small.

On the other hand, as shown on the right side of FIG. 3, if the pixel level changes slowly within block 502 and the brightness or hue gradually increases or decreases in a fixed direction, as shown in FIG. 4(B), Each pixel level in block 502 is the minimum pixel level M
From IN L to the maximum pixel level MAX-L, the pixel levels are distributed in a --like manner.

Therefore, when each pixel level within block 502 is sorted into three level regions as shown in FIG. 3, there will be a large number of pixels that fall into the intermediate level range.

The value Nih compared with the number of pixels belonging to this intermediate level range, n, reliably distinguishes between the steep pattern shown on the left side of Figure 3 and the gentle pattern = 20- shown on the right side of the same figure. The mode selection circuit 208 selects mode B for the steep pattern shown on the left side of FIG. 3, and mode C for the slow pattern shown on the right side of FIG.

That is, the maximum pixel level MA within acquisition block 502
The difference between X-L and the minimum pixel level MIN L is 18.
As described above, when the pixel levels are distributed over a wide level range, two block representative levels are determined by the representative gradation determining circuit 210 when the number of pixels belonging to the intermediate level range is small.

Furthermore, when the number of pixels belonging to the intermediate level range is large, four block representative levels are determined.

As explained above, according to this embodiment, even if the pixel levels in the captured block 502 are distributed over a wide range, there are two representative levels when the block 502 includes an outline or the like. As it is decided,
Resolution component φ1. The amount of code for φ2 can be reduced.

Therefore, block encoding of multilevel images can be performed more efficiently.
It becomes possible to perform the process at high speed.

In addition, when approximating contour lines etc. at many levels, 2
Approximation at two levels allows for clearer restoration, which also makes it possible to improve image quality.

[Effects of the Invention] As explained above, according to the present invention, even if the pixel levels within a block are distributed over a wide level range, when the distribution has a bimodal characteristic, the pixel levels within the current block are contains contour lines, etc., so it is possible to reduce the amount of encoded data by approximating it using a large number of representative levels, or by approximating it using a small number of levels.

Therefore, it is possible to perform block encoding of a multivalued image efficiently (at high speed).

Furthermore, according to the present invention, the more a contour line or the like is approximated by a smaller number of representative levels, the more clearly the contour line is restored, and therefore the image quality can be further improved.

[Brief explanation of the drawing]

Fig. 1 is an explanatory diagram of the principle of the invention, Fig. 2 is an explanatory diagram of the block configuration of the embodiment, Fig. 3 is an explanatory diagram of the quantization effect at the pixel level, Fig. 4 is an explanatory diagram of the frequency distribution at the pixel level, and Fig. 5 is an explanatory diagram of the pixel level frequency distribution. is an explanatory diagram of the block encoding process,
FIG. 6 is an explanatory diagram of gradation change characteristics. 200...e buffer memory, 202...gradation change detection circuit, 204@@+ conversion circuit, 206.e.state detection circuit, 208...mode identification circuit, 210--.representative tone determination Circuit, 212・Contest・Memory circuit, 214・・Sha comparison circuit, 216.218@・・Buffer memory, 220・・φ encoding circuit, 222.224・φφ memo 1ha 226・Cloud fist signal synthesizer, 228 - Difference value generation circuit, 230... Difference value encoding circuit, 232 Reference/write buffer memory, 234/Fist/M standard bell generation circuit, 236... Reference level encoding circuit, 238... buffer memory. Wa resistanceless light and darkness Be Δ<ma reference level La La level interval Ld block

Claims (1)

[Claims] Means (10) for capturing an image formed by pixels of a multilevel representation block by block; means (12) for instructing the generation of a pixel level that is representative of the block; and means (14) for instructing the generation of a small number of pixel levels representative of the block when the pixel levels of the captured block are distributed over a wide range and there are few intermediate level pixels; ), means (16) for instructing the generation of a large number of pixel levels representative of the block when the pixel levels of the captured block are distributed over a wide range and there are many intermediate level pixels; means (18) for generating a level based on each pixel level of the acquired block; means (20) for quantizing each pixel level of the acquired block into one of the generated block representative levels; and block representative levels. means (22) for determining the distribution width of the captured block; and means (24) for determining the standard pixel level of the captured block.
); and means (26) for encoding pixel level quantization content, block representative level distribution width, and standard pixel level.