JP3066278B2 - Image encoding device and image decoding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coding device and a decoding device for image data. In particular, the present invention relates to an image encoding device capable of encoding at a high compression rate based on image contour information.

[0002]

2. Description of the Related Art Conventionally, various methods have been proposed for encoding image data at a high compression rate. Many of such methods attempt to compress using the statistical properties of images, but in recent years, methods of performing encoding using visual features when viewing image data are being proposed. .

[0003] Most of the image data is a two-dimensional signal obtained by projecting a real three-dimensional world onto a plane by a camera mechanism, and includes a structure that appeals to the sight of a person closely related to the real three-dimensional world. Is reflected. In such a method, the structural features of the image are positively used during encoding. Image coding techniques from such a standpoint include edge,
A method that uses the two-dimensional structure of the image such as contours and regions,
A scheme using three-dimensional motion to be encoded and other schemes have been studied.

[0004] In particular, a technique of performing encoding by extracting a two-dimensional structure is considered to be effective as encoding for a still image.

[0005] The two-dimensional structure extraction coding method is classified into a method using contours and edge components and a method using integrated area information. In particular, a method of extracting only edges and contours and encoding it as a binary line drawing is a method that has long been studied for the purpose of recognizing the shape of an object. When using this technique for image coding,
In many cases, the ultimate purpose is not to represent an image as a mere line drawing, but to reproduce it as a natural image having a luminance gradation.

[0006] Among the encodings based on two-dimensional features, some proposals have been made regarding a method using a contour line. For example, “Image Information Compression”, Hiroshi Harashima, 24
Several encoding methods based on the two-dimensional structure are introduced on pages 3 to 245. For example, a code algorithm of Sketch Based Coding by Carlsson is introduced as a representative one introduced in this document. According to this encoding algorithm, first, contour information is generated from the zero crossings of the original image data using a Laplacian-Gaussian operator. Then, based on the contour information generated using the Laplacian-Gaussian operator, the contour shape and the luminance component on the contour are encoded. In this way, encoding is performed only on the contour, and is transmitted through a communication path or stored in a predetermined storage unit.

To decode the data encoded in this way, first, the contour shape and the luminance component on the contour are decoded based on the encoded data. That is, if the encoded image data is decoded as it is, only the contour shape and its luminance component can be obtained. Then, points other than the contour are interpolated by the variation minimum standard. In addition,
If necessary, sub-sampled data of the coding residuals may be added to improve the interpolation quality.

According to Carlsson's experiment on a still image, it has been confirmed that an image which can sufficiently perceive an object can be reproduced even if the compression ratio is about one-tenth to one-hundredth.

[0009] This Sketch Based Codi
FIG. 5 shows the state of encoding by the ng encoding algorithm.
(A) to (d) are shown. Here, FIG. 5A shows original image data to be encoded. FIG. 5B shows the result of obtaining the outline information from the zero crossing of the original image data by using the Laplacian-Gaussian operator as described above. The contour information as shown in FIG. 5B is called so-called sketch data.

[0010] Encoding is performed on the contour information shown in Fig. 5B, and examples of decoding the encoded data are shown in Figs. 5C and 5D. I have. The example shown in FIG. 5C is 0.11 bit /
pel, that is, an example of a decoded image when a point other than the contour is obtained from the data on the contour by interpolation at a compression ratio of 1/75. Since the pixel values are reproduced by interpolation in portions other than the outline, subtle shading is lost, but the shape is reproduced almost exactly, so that it is easy to visually identify the target. In the example shown in FIG. 5D, data obtained by sub-sampling the coding residual by 16 × 16 pixels is added as coded data and transmitted to the decoding side. This is an example of a case where decoding is performed based on data. In this case, 0.13 bit /
The data amount is about pel, that is, the compression ratio is 1/63.

[0011]

As described above, if encoding is performed based on a two-dimensional structure of image data, for example, based on a contour, a high compression ratio can be obtained without significantly deteriorating visual image quality. Encoding was possible.

However, with the recent development of communication and broadcasting, the amount of image data to be handled is rapidly increasing.
Along with this, image coding with a higher compression rate is desired, and the image compression rate can be changed according to the application based on the required image quality depending on the application field, for example, the reproduction method. It is strongly desired to do so.

According to the above-described method of encoding information relating to contours, encoding with a very high compression rate can be performed. On the other hand, however, the ratio of the extracted contour portion to the entire image data is determined. Whether or not it depends on the image. Therefore, for example, when the amount of information that can be transmitted is limited, such as in a videophone, the inconvenience that the transmission time differs for each image occurs.

SUMMARY OF THE INVENTION The present invention has been made in view of such a problem, and an object of the present invention is to provide an image data encoding apparatus which is an encoding method based on contour information and which can easily change a compression ratio. To provide. It is another object of the present invention to construct coded data hierarchically so that a receiving side or a decoding side can freely select image data having a desired accuracy.

[0015]

According to a first aspect of the present invention, there is provided edge data extracting means for extracting edge data of image data, and edge data extracted by the edge data extracting means. Encoding means for encoding, wherein the edge data extracting means quantizes a pixel value of each pixel of the image data to a predetermined N (N is an integer of 2 or more) level, and A quantizing means for calculating a value, and a pixel in the image data, wherein a pixel whose quantized pixel value is different from a quantized pixel value of an adjacent pixel is determined to be edge data. And an edge data determination unit for performing the above operation, wherein the number N of quantization levels in the quantization unit is variable.

[0016]

According to a second aspect of the present invention, there is provided edge data for image data.
Edge data extraction means for extracting the edge data;
Encoding means for encoding the edge data extracted by the data extracting means, wherein the edge data extracting means calculates a pixel value of each pixel of the image data. Quantizing means for quantizing to a predetermined N (N is an integer of 2 or more) level to calculate a quantized pixel value, and J (where J is an integer of 2 or more) N Hierarchical quantization control means for supplying different numbers N1 to NJ and causing the quantization means to calculate J kinds of the quantized pixel values for one piece of the image data. The data extracting means extracts a plurality of pieces of edge data having different densities from one piece of the image data.

According to a third aspect of the present invention, in the image encoding apparatus according to the second aspect of the present invention, the encoding means changes the edge data from the coarsest edge data for the N having the smallest value to a more dense edge data. , Hierarchical encoding means for sequentially performing encoding and outputting the sequentially encoded edge data to the outside;
It is characterized by including.

According to a fourth aspect of the present invention, in the image encoding apparatus according to the third aspect of the present invention, the hierarchical encoding means includes the edge data to be output already contained in the coarser edge data already output. If the data is included, only the remaining edge data excluding the already included edge data is encoded and output.

According to a fifth aspect of the present invention, there is provided an image decoding apparatus for decoding image data encoded by a third or fourth image encoding apparatus. Are sequentially decoded, and the low-quality decoded image data is
Hierarchical decoding means for decoding up to high quality decoded image data in order.

According to a sixth aspect of the present invention, in the image decoding apparatus according to the fifth aspect of the present invention, the hierarchical decoding means is arranged so that the decoding time or the information amount of the image to be decoded is limited.
If the limit is exceeded during the decoding of the image data in order, the hierarchical decoding means stops the decoding and updates the latest data obtained up to the time of the stop. Is output as final decoded image data.

[0022]

The edge data detecting means according to the first aspect of the present invention determines that pixels having different quantized pixel values are edge data. Therefore, whether or not the data is edge data can be quickly and easily determined, and the number N of quantization levels is variable. Therefore, the compression ratio can be adjusted according to the quality required for the image data.

[0023]

The hierarchical quantization control means according to the second aspect of the present invention supplies a plurality of quantization levels N to a plurality of quantization means to extract a plurality of edge data. Therefore, a plurality of edge data having different densities can be obtained. Therefore, it is possible to arbitrarily select image data of a desired quality according to the use of the image data. In the third aspect of the present invention, encoding and output are sequentially performed from hierarchically encoded coarse edge data to dense edge data. Therefore, if decoding is performed in order from the rough edge data on the decoding side, it is possible to obtain a plurality of types of image data in order from low-quality image data to high-quality image data.

According to the fourth aspect of the present invention, when the edge data to be output is already included in the coarse edge data preceding the edge data, the remaining edge data whose part is omitted is decoded. And output. Therefore, it is possible to prevent the encoding and output of the duplicated data, and it is possible to reduce the encoding time and the amount of transmission.

According to the fifth hierarchical decoding means of the present invention, decoding is performed from low-quality decoded image data to high-quality decoded image data in order. Therefore, it is possible to arbitrarily select image data of required image quality.

The hierarchical decoding means according to the sixth aspect of the present invention comprises:
If there is a certain restriction while decoding from low-quality decoded image data to high-quality decoded image data in order, decoding is stopped and the most Outputs high-quality decoded image data. Therefore, even when the amount of edge data differs for each image, it is possible to decode the image data at regular time intervals or obtain image data of a constant quality (a constant information amount). Becomes

[0028]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of an image encoding apparatus and an image decoding apparatus according to a preferred embodiment of the present invention. As shown in FIG. 1A, image data to be encoded is first supplied to a quantization unit 10 and an edge data determination unit 12. The quantization means 10 classifies each pixel value of the input image data based on a certain threshold value, and reduces the number of possible pixel values. For example, as image data, 25
Six-gradation image data is often used.
That is, six gradations are reduced to four gradations. Such quantization is useful for detecting edge data in image data. The edge data determination means 12
Observe the pixel value quantized by the quantization means 10,
A pixel having a different pixel value from an adjacent pixel value in the quantized pixel value is determined to be edge data. That is, in the image encoding device according to the present embodiment, a portion where the value of the pixel value is largely changed is determined to be a pixel included in the edge data. As described above, the edge data is detected very easily by simply selecting a pixel whose pixel value is different from an adjacent pixel value while the number of gradations is reduced by the quantization means 10. It becomes possible. In the conventional image encoding apparatus, a Laplacian-Gaussian operator or the like must be used to determine whether or not the data is edge data, as described above, and a complicated calculation is required. On the other hand, according to the present embodiment, it is sufficient to simply compare the values, so that higher-speed image encoding can be performed.

Further, it is preferable that the quantization means 10 quantizes, for example, to two levels or eight levels in addition to the four levels. Quantizing means 1 in the present embodiment
0 performs quantization based on the number N of quantization levels supplied from the hierarchical quantization control means as shown in FIG. Therefore, it is possible to respond to each request by setting the number of quantization levels N to an appropriate value according to the image quality, the amount of image data information, and the like required according to the application.

The edge data determined in this way is supplied from the edge data determining means 12 to the encoding means 16. The encoding means 16 has a configuration similar to that of the above-described conventional image encoding apparatus, encodes only edge data, and outputs encoded image data.

FIG. 1B is a block diagram showing the configuration of an image decoding apparatus that decodes the encoded image data encoded as described above and outputs the original image data. The configuration of the image decoding device shown here is almost the same as that of a conventional image decoding device. That is, the encoded image data is first decoded by the decoding means 18 to obtain the original pixel value. Next, the pixel values within the edge data portion are interpolated by the interpolation means 20 and reproduced. For this interpolation, various methods such as linear interpolation and spline interpolation can be used as in the above-described conventional decoding apparatus.

As described above, according to the present embodiment, in extracting edge data, each pixel value of image data to be encoded is once quantized, and edge data is determined based on the quantized pixel value. . Therefore, it is possible to easily extract the edge data, and it is possible to perform the image encoding with a desired accuracy by changing the number of quantization levels N.

The hierarchical quantization control means 14 normally supplies one N to one image data, but it is also preferable to apply a plurality of quantization levels N to one image data. . That is, quantization based on different numbers of quantization levels is performed on one image data in the quantization means 10. Since the edge data is determined based on the different quantizations, image encoding with various precisions is performed. In this case, the encoding unit 16 sequentially outputs data having a low encoding accuracy to data having a high encoding accuracy. As described above, the effect of sequentially outputting encoded data of a plurality of precisions appears when an image is decoded. That is, although the amount of edge data differs depending on each image data, it is often desired to suppress the data amount for one image data to a constant value. As described above, in order to keep the information amount of image data to a constant value, coded data with low accuracy is stored in order, and storage of image data is terminated when a predetermined storage area is full. Then, although the accuracy differs slightly depending on each image data, the information amount of one image data can be set to a constant value, which can contribute to the management of the image data.

In other cases, for example, the number of frame images that can be sent per second is more important than video quality in videophones and the like. Furthermore, in a videophone or the like, the capacity of a communication line is generally very limited in many cases. Even in such a case, by outputting coded image data with high accuracy and sequentially outputting coded image data with high accuracy, it is possible to make full use of the capacity of the communication line, Can be maintained at a constant value.

Further, when the coded image data having a high precision is sequentially transmitted from the coded image data having a low precision, the data of the coded image data having a high precision and the coded image data having a high precision are transmitted. Is preferably not sent. For example, a pixel that is extracted as edge data when quantized with the number of quantization levels 2 is always edge data when quantized with the quantization level number 4, and is thus obtained, for example, with the quantization level number 2. Encoded image data and the number of quantization levels 4
It is preferable not to send the common part (pixel) with the coded image data obtained in (1) when transmitting the coded image data obtained with the quantization level number 4. That is, when the coded image data obtained with the quantization level number 4 is transmitted after the coded image data obtained with the quantization level number 2 is transmitted, the coded image data newly obtained with the quantization level number 4 is transmitted. Only the encoded image data thus transmitted is transmitted. By employing such a configuration, it is possible to efficiently encode image data.

The coded image data hierarchized in this way has various other applications. For example, when the number of pixels of the image data is extremely large, the data amount of the encoded data obtained by encoding the image data may be large. In such a case, decoding may also take considerable time. Therefore, it is impossible for the conventional image decoding apparatus to cope with the case where it is desired to quickly know only the general state of the original image data. However, the image encoding device according to the present embodiment,
According to the image decoding apparatus, the image data is first transmitted from coarse-accuracy image data, so that it is possible to inform the operator of the rough state of the image. Such uses may be used for cataloging image databases. For example,
It is considered that it is often the case that only the image data representing a simple state is sufficient instead of the exact original image data when searching for the target image data. In particular, when searching for image data, the decoding speed is considered to be more important than the accuracy of the image data.

Next, how the edge data is extracted according to the present embodiment and how the original image is restored from the edge data will be described with reference to an explanatory photograph. FIG. 2 shows a claire which is generally used as a standard image. This image cla
FIG. 3A illustrates a pixel obtained by quantizing ire using the quantization level number 4 and then extracting a pixel whose quantized pixel value is different from an adjacent pixel value. For example, in the example shown in FIG. 3A, the number of pixels extracted as edge data is about 17% of the total number of pixels of the image data. That is, the information amount of the image data is reduced to about 1/6 merely by the operation of simply extracting the edge data. Only the pixels of the edge data obtained in this way are encoded to generate encoded image data. FIG. 3 is a diagram in which an original image is restored based on the encoded image data obtained in this manner.
This is shown in (b). In the image shown in FIG. 3B, pixels belonging to portions other than the edge data are calculated by interpolation from the pixel values of the pixels belonging to the edge data, similarly to the above-described conventional image decoding apparatus. I have. As a result, although the gradual changes in color and brightness have been lost,
It will be understood that the shape of the person is almost completely restored. FIG. 4 shows an example in which the number of quantization levels N is 2. FIG. 4A shows the state of the edge data extracted when the number of quantization levels N is two. When the number N of quantization levels is 2, the number of pixels extracted as edge data is approximately 13% of the total number of pixels of image data. FIG. 4B shows an example in which the values of pixels other than the edge data are reproduced by interpolation from the edge data thus extracted, and the original image is restored. Figure 3 shows the image quality.
Although deteriorated compared to the example shown in (b),
The shape is almost exactly conveyed.

As described above, according to this embodiment, it is possible to easily perform edge data of image data by quantizing each pixel value constituting the image data. Therefore, it is possible to quickly extract edge data, and it is possible to obtain an image encoding device capable of quickly processing a large amount of image data.

Further, by changing the number N of quantization levels, it is possible to encode images with various precisions.

Then, hierarchical quantization is performed for various numbers of image levels N, edge data corresponding to each quantization is extracted, and a plurality of edge data having different degrees of density are extracted. Then, since encoded image data with coarse accuracy to encoded image data with high accuracy are sequentially output, image data with desired accuracy can be obtained on the decoding side, and when communication capacity and decoding time are limited. Also in such cases, there is an effect that image encoding / decoding can be performed while effectively utilizing the communication capacity and the decoding time.

When a plurality of pieces of edge data having different degrees of density are extracted and the encoded image data is transmitted, high-precision data which is included in the coarse-precision data is used. If data is not sent when sending high-precision data, efficient use of communication capacity,
Alternatively, the storage capacity can be efficiently used.

[0043]

As described above, according to the first aspect of the present invention, since the edge data can be easily extracted, there is provided an image encoding apparatus capable of encoding a large amount of image data quickly. Since the number of quantization levels N is variable, it is possible to perform image coding with accuracy corresponding to various uses.

[0044]

According to the second aspect of the present invention, since one image data is subjected to image encoding with various precisions, the decoding side can arbitrarily select encoded image data with a desired precision. Becomes possible.

According to the third aspect of the present invention, since encoded image data having coarse accuracy is sequentially output to encoded image data having high accuracy, a rough state of the image is first obtained on the decoding side. It is possible to sequentially generate image data with high accuracy.

According to the fourth aspect of the present invention, pixels which are included in high-precision coded image data and which are also included in coarse-precision coded image data already transmitted earlier. Is not sent again when sending high-precision image data. As a result, more efficient generation and transmission of encoded image data can be achieved.

According to the fifth aspect of the present invention, there is provided an image decoding apparatus for decoding image data encoded by the image encoding apparatus according to the third or fourth aspect. Alternatively, an effect similar to that of the fourth invention is exerted.

According to the sixth aspect of the present invention, when coded image data of a plurality of precisions are sequentially transmitted, and when there is a limit on the decoding time or the amount of information of the image, the limit is fully satisfied. Since image decoding when used is obtained, time or the amount of information can be effectively used.

[Brief description of the drawings]

FIG. 1 is a configuration block diagram of an image encoding device and an image decoding device according to a preferred embodiment of the present invention.

FIG. 2 is an explanatory photograph of a halftone image in which an original image of the image used in FIGS. 3 and 4 is displayed on a display .

[Figure 3] Disupu edge data and reconstructed image from the edge data when the number of quantization levels N is 4
5 is an explanatory photograph of a halftone image displayed on a ray .

FIG. 4 shows edge data obtained when the number of quantization levels N is 2, and an image restored from the edge data.
4 is an explanatory photograph of a halftone image displayed on a display .

[5] The manner of encoding by the conventional encoding device Disconnect
It is an explanatory photograph of a halftone image displayed on play .

[Explanation of symbols]

Reference Signs List 10 quantization means, 12 edge data determination means, 14
Hierarchical quantization control means, 16 encoding means, 18 decoding means, 20 interpolation means.

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int. Cl. ⁷ , DB name) G06T 9/00-9/40 H04N 1/41-1/419 H04N ^7/ 24-7/68

Claims (6)

(57) [Claims] 1. Edge data extracting means for extracting edge data of image data; encoding means for encoding edge data extracted by the edge data extracting means;
Wherein the edge data extracting means quantizes a pixel value of each pixel of the image data to a predetermined N (N is an integer of 2 or more) level, and calculates a quantized pixel value. A pixel in the image data, wherein the quantized pixel value of the pixel is different from the quantized pixel value of an adjacent pixel;
An image encoding apparatus comprising: edge data determining means for determining edge data; wherein the number of quantization levels N in said quantization means is variable. 2. Edge data extracting means for extracting edge data of image data; encoding means for encoding edge data extracted by the edge data extracting means;
Wherein the edge data extracting means converts a pixel value of each pixel of the image data into a predetermined N (N is 2
A quantizing means for quantizing to the above-mentioned integer level and calculating a quantized pixel value; and supplying N (J is an integer of 2 or more) different numbers N1 to NJ to the quantizing means. And hierarchical quantization control means for causing the quantization means to calculate the J kinds of the quantized pixel values for one piece of the image data, wherein the edge data extraction means comprises one piece of image data. An image encoding apparatus for extracting a plurality of edge data having different densities from the image data. 3. The image encoding apparatus according to claim 2, wherein said encoding means converts the coarsest edge data for the N having the smallest value into a denser edge data.
Encoding is performed in order, and the edge data is encoded in order.
An image encoding apparatus, comprising: a hierarchical encoding unit that outputs data to the outside. 4. An image coding apparatus according to claim 3, wherein
If the edge data to be output is already included in the coarser edge data already output, the hierarchical encoding means may delete the remaining edge data excluding the already included edge data. An image encoding apparatus for encoding and outputting only data. 5. An image decoding apparatus for decoding image data encoded by the image encoding apparatus according to claim 3 or 4, wherein said edge data inputted first is sequentially decoded. From low-quality decoded image data to high-quality decoded image data.
An image decoding device, comprising: a hierarchical decoding unit that sequentially decodes data up to the data. 6. The image decoding apparatus according to claim 5, wherein the hierarchical decoding means is configured to perform the decoding when the decoding time or the information amount of the image to be decoded is limited. If the above-mentioned limit is exceeded while decoding image data in order, the decoding is stopped and the latest decoded image data obtained up to the time of the stop is stopped. An image decoding device for outputting the decoded data as final decoded image data.