JPH10126621A - Image compressing method/device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve a compression rate as a whole while the image quality of a necessary part is maintained by obtaining hues and saturations for the respective image areas constituting an image and executing encoding at the compression rate decided based on them. SOLUTION: In the process of the encoding of JPEG, a prescribed preprocessing 1 is added before quantization. The preprocessing 1 is a processing for converting the respective areas of the compressed image based on a function decided based on the hues and the saturations or a table. The compression rate of a block which is preprocessed by the table whose value of K for regulating the quantization step of the function is large becomes high and the block which is preprocessed by the table whose value of K is small becomes low. The respective image areas constituting the image us encoded by the compression rate decided based on the hues and the saturations of the image areas. Thus, image quality is maintained on a part whose image quality is concerned and the compression rate of a part whose image quality is not concerned is improved. Then, the compression rate as a whole can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression method and apparatus for compressing image information representing an image by encoding the image information, and more particularly to a method and apparatus for compressing a color still image such as a photographic image.

[0002]

2. Description of the Related Art Conventionally, orthogonal transformation, quantization, entropy coding and the like have been known as image compression techniques. The orthogonal transform focuses on the fact that image information having a high correlation is concentrated on the low frequency region when viewed on the frequency axis. This is a compression technique that reduces the amount of information as a whole by encoding. Quantization is a compression technique that divides image information by a certain numerical value (quantization step) and reduces the number of bits representing the image information by approximation at a limited number of discrete levels. In addition, entropy coding is a compression technique that reduces the total code amount by assigning a short code to a value with a high occurrence frequency and a long code to a value with a low occurrence frequency when assigning a binary code to image information. is there.

[0003] JPEG, which is widely known as a color still image encoding standard, encodes image information by combining the above-mentioned compression techniques. Basically, as shown in FIG. Is divided into image blocks of, for example, 8 × 8 pixels, and a discrete cosine transform (DCT), which is one of orthogonal transforms, is performed on each image block.
The image information is compressed by subjecting the transform coefficients obtained by the above to quantization and further subjecting the quantized values to Huffman encoding or the like, which is entropy encoding. Further, when an image is reproduced, the reverse procedure, namely, entropy decoding, inverse quantization, inverse discrete cosine transform (IDC)
T) decompresses the compressed image information.

[0004] JPEG is relatively easy to obtain as hardware or software for both encoders and decoders. In addition, the degree of image quality degradation due to encoding is changed by adjusting parameters, and the speed of decoding is increased. Because of the convenience of being able to change it, it is a technology that is widely used not only in the industry but also in general household personal computers.

[0005]

In the above-described image compression technique, it is inevitable that the image quality will deteriorate to some extent if the compression ratio is increased, and the compression ratio is determined by the allowable range of the image quality deterioration. It is believed to be something.

In particular, in the above-mentioned JPEG, the compression ratio can be adjusted only for each image to be compressed. For this reason, for example, a portion of a main subject whose high image quality is to be maintained, such as a photographic image, and a certain degree of image quality deterioration are allowed. In an image composed of a background image to be processed, the compression ratio has to be determined according to the image quality required for the main subject. In other words, if there is even a small portion of the image where high image quality is desired to be maintained, the compression ratio of the entire image is determined according to the allowable range of image quality deterioration of that portion. could not.

SUMMARY OF THE INVENTION In view of the above problems, it is an object of the present invention to provide an image compression method and apparatus capable of maintaining the image quality of a necessary portion and increasing the overall compression ratio as much as possible. Things. More specifically,
When a human distinguishes a main part from a non-principal part in an image, he or she distinguishes based on hue and saturation. Focusing on the fact, while maintaining a high compression rate as a whole, an image compression method that can maintain a relatively high image quality for a main part in an image or a color portion where image quality deterioration is strongly felt, and It is intended to provide a device.

[0008]

SUMMARY OF THE INVENTION An image compression method and apparatus according to the present invention compress image information by encoding image information representing an image. Each of the plurality of image regions constituting the image is compressed. , And encodes image information included in each image area at a compression rate determined based on the hue and saturation.

[0009] Here, "hue and saturation" are three colors.
Hue and saturation among the hue, saturation, and lightness that are the elements. Further, "determining hue and saturation for each of a plurality of image regions constituting the image" means that after sequentially calculating hue and saturation for all pixels, the image is divided into a plurality of regions and divided into each region. The average of the hue and saturation of the included pixels may be obtained, or the image may be divided into a plurality of regions first,
For example, the hue and saturation of the representative pixel may be obtained for each region, or the average hue and saturation of the pixels included in the region may be obtained.

[0010] At this time, the "image area" is large enough that the values obtained as the hue and saturation of the area are appropriate values indicating the hue and saturation visually recognized on the actual image. Desirably. For example, when it is desired to maintain the image quality of a skin color portion, that is, the entire face of a subject in a snapshot or the like, it is preferable to obtain a hue and saturation using an area sufficiently smaller than the size corresponding to the face of the subject as an image area.

The "compression rate determined based on the hue and saturation" means that the compression rate is reduced when the hue and saturation indicate a color in an area where high image quality is desired to be maintained. It is routinely known that humans distinguish between main parts and non-main parts in an image based on hue and saturation. If so, the human being, which is the main part, and the background trees, which are the non-main parts, are distinguished by the skin color, which is usually the color of clothes or the color of the face. In other words, the present invention intends to lower the compression ratio of the main part in an image by utilizing this human visual characteristic.

As a specific method for “encoding at a compression rate determined based on the hue and saturation”, for example, if the encoding performs quantization of image information, There is a method of determining the compression ratio by determining the value of the quantization step of quantization based on the hue and saturation, and performing encoding. Alternatively, in a process other than the quantization of encoding, a process may be performed in consideration of hue and saturation.

As a general method of performing quantization in the compression process, image information is divided into a plurality of image blocks before quantization, and this image information is subjected to an orthogonal transform for each image block. And a method of performing quantization on the transform coefficient obtained by the orthogonal transform is known,
Also in the present invention, the quantization based on the hue and the saturation may be performed after the orthogonal transformation is performed.

In this case, the above-mentioned hue and saturation may be obtained for each image block which is a unit of the orthogonal transformation.

[0015] A method of performing entropy coding on quantized image information is generally known as in the case of orthogonal transform. In the present invention, quantization based on the above hue and saturation is performed. After that, such entropy coding may be further performed.

Further, the quantization is firstly subjected to a pre-processing for determining a compression ratio based on a function or a table determined on the basis of hue and saturation. It may be realized by performing quantization in the technology and adding preprocessing to the conventional technology.

In general, in a photographic image or the like, image quality deterioration of a face portion of a main subject is particularly worrisome. Therefore, when hue and saturation indicate flesh color, the compression ratio should be reduced. desirable.

[0018]

According to the image compression method and apparatus of the present invention, when a human distinguishes a main part from a non-principal part in an image based on hue and saturation, or human visual Focusing on the fact that there is a color that strongly senses image quality degradation and a color that does not feel very much as a characteristic, each image region that composes an image is compressed at a compression ratio determined based on the hue and saturation of the image region. Since each is encoded, the image quality is maintained for the part where the image quality is not a concern, while the compression rate can be increased as a whole by increasing the compression ratio of the part where the image quality is not a concern. Also, for example, when it is desired to intentionally reduce the image quality of only the human face part for privacy protection, according to the present invention, for example, the compression ratio is increased only for the skin color area, and the face part is maintained while maintaining the surrounding image quality. Only can be blurred.

Here, if the compression rate is determined by performing quantization as the encoding and determining the value of the quantization step of the quantization based on the hue and saturation, for example, Focusing on a certain color in the degree space, the quantization step can be changed little by little according to the distance from that color (the area that represents the color in the hue / saturation space). The rate can be adjusted. For example, in a method of changing a compression rate by changing a compression method based on hue and saturation, hue and saturation must be divided into a limited number of groups (for example, the number of compression methods) in advance. It is not possible to adjust the compression ratio, and the effect of finely setting the compression ratio by adjusting the quantization step has a great effect.

Further, as in the case of JPEG, for example, if the quantization is performed after performing an orthogonal transformation such as DCT, or the entropy coding is further performed after the quantization, the overall compression ratio can be further improved.

In this case, it is necessary to block the image in order to perform the orthogonal transformation. Therefore, if the image area is defined as this block and the hue and saturation are obtained in block units,
The image dividing process can be combined into one step, which is more preferable in the configuration of the apparatus.

Also, while keeping the JPEG quantizer as it is,
Before that, add a conversion process based on hue and saturation and add 2
If the quantization is performed in stages, the load at the time of mounting is reduced.

In general, image compression is often performed not only when images are stored but also when they are transferred and distributed. for that reason,
In order for the image compression technique to be widely used, it is assumed that the user at the transfer destination has decoding means. Therefore, in a situation where a standard such as the above-mentioned JPEG has been established as an image compression technique, a special decoding means is required even if a technique superior to JPEG in terms of image quality and compression ratio is developed. If the technology does, it may not be acceptable to general users. Alternatively, it takes a long time and effort to standardize the technology.

On the other hand, the method and apparatus for determining the value of the quantization step of the quantization based on the hue and the saturation among the method and apparatus of the present invention, so to speak, only adjust the fineness of the quantization. The coding procedure does not change. Therefore, the image compressed by the image compression method of the present invention can be reproduced by the same decoding procedure as the conventional one, and does not require any new decoding means. That is, for example, if a JPEG-compliant encoding means is improved on the side that provides a photographic image or the like as in the present invention, a general user who uses the image can improve the image quality only by reproducing the image in the same manner as before. The effect of improving the compression ratio can be obtained.

That is, it can be said that the present invention not only has a technical effect, but also is a highly utilizable invention in accordance with the trend of the actual industry.

[0026]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an image compression method and apparatus according to the present invention will be described with reference to the drawings. In the following description, an embodiment as an improvement of JPEG, which is an industry standard, will be described. However, as described above, the feature of the present invention is that humans can color main and non-main parts of an image by color. In order to reduce the compression ratio of the main part, image compression is performed at a different compression ratio for each region based on hue and saturation in order to reduce the compression ratio of the main part. It does not limit the technical scope of the. However, since the field of image compression is being standardized, it is actually necessary to implement the features of the present invention in JPEG and other standard technologies, or technologies to be standardized in the future, as in the present embodiment. Is desirable.

FIG. 1 is a block diagram showing an outline of an embodiment of the present invention. As shown in this figure, the present embodiment is obtained by adding a predetermined preprocessing 1 before quantization in the above-described JPEG (FIG. 4) encoding process.
The decoding process is no different from the conventional one. At the time of inverse quantization, the table used for quantization is used, and the image after the preprocessing at the time of encoding is reproduced.

Here, the pre-processing is a process of converting each area of the image to be compressed based on a function or a table determined based on the hue and saturation of the area. As shown in FIG. 2, in JPEG, an image is divided into 8 × 8 blocks, and DCT is applied to each block. In this embodiment, this block is used as a range for calculating hue and saturation. I have.

The hue and saturation may be obtained by averaging the hue and saturation of 8.times.8 pixels included in each block. For example, one or several pixels at the center may be set as a representative pixel. The hue and saturation of the pixel may be used as the hue and saturation of the block. Further, the obtained hue and saturation may be held as they are, or, for example, focusing on a specific color called flesh color, assuming that the flesh color is 10, and as the distance from the flesh color area on the hue / saturation space increases, 9,
It may be stored as information having a small value such as 8 or 7. Alternatively, the information may be simply converted into binary information such as 1 for a skin color and 0 otherwise, and stored.

The hue and saturation obtained and held in this way, or information relating thereto, is referred to when a function or a table is determined in the preprocessing 1. Preprocessing 1 of the present embodiment is a process of transforming a transform coefficient obtained as a result of DCT by a function having a shape as shown in FIG. 3, and a value K defining a quantization step of the function is a so-called quantizer. Similarly to the above, the setting is made for each DCT coefficient position (the position of the DCT transform coefficient in the frequency space). A plurality of sets of such setting values are prepared as a pre-processing table, and the plurality of tables are switched at any time according to the hue and saturation of the block to be processed.

Here, the pre-processing 1 is a process for quantizing image information broadly like a JPEG quantizer. However, since the main purpose is not to adjust the quantization itself but to adjust the compression ratio, The value is not necessarily a continuous integer value (index). Here, as the value of K increases, the pre-processed image information becomes a limited type of value, so that the efficiency of the subsequent quantization is improved, and the entropy of the image information is reduced. The encoding efficiency is also improved.
That is, a block pre-processed by a table having a large value of K has a high compression ratio (large image quality deterioration), and a block pre-processed by a table having a small value of K has a low compression ratio (less image quality deterioration). ).

Therefore, in this embodiment, considering that the skin color is most visually sensitive, the K of all the coefficient positions in the table of the preprocessing 1 is maintained for the skin color in order to maintain the accuracy.
Is set to 1. Gray is also a color in which image quality degradation is relatively conspicuous, and the values of K at all coefficient positions are, for example, 2
And so on. For other colors, no particular precision is required, and the values of K at all coefficient positions are set to larger values. As described above, three types of different preprocessing tables are prepared in advance according to the colors, and the tables are switched according to the hue and saturation of the block to be processed.

Alternatively, a plurality of types of skin color tables and a plurality of types of gray tables are prepared. For example, when the skin color itself is used, the value of K is 1, and the color of the image block to be compressed is the skin color in the hue saturation space. As the distance from the region increases, the value of K becomes 2,3. . . By switching to the table, compression according to the characteristics of the image can be performed more accurately.

Further, instead of changing the value of K and switching the table prepared in advance, an appropriate value of K may be obtained in accordance with the hue and saturation, and the processing may be performed.

When the compression is performed in this way, the compression ratio of the whole image is sufficiently ensured because the accuracy of the parts other than the skin color and the gray is reduced. Further, when this image is decoded, the flesh color or gray portion is reproduced with high accuracy, so that even if the accuracy of the other portions is reduced, the image quality is hardly deteriorated as a whole.

It should be noted that the value of K in the preprocessing table does not necessarily need to be the same at all coefficient positions as in a general quantization table. For example, if the user wants to maintain the accuracy of the skin color for the low-frequency component but does not need particularly high accuracy for the high-frequency component, the value of K at the coefficient position corresponding to the low-frequency component is set to 1 and the The value of K at the coefficient position may be set to a relatively large value.

In the above description, quantization is performed by two-stage processing of pre-processing and a conventional quantizer in order to clarify the features of the present invention. However, processing may be performed simultaneously by one table. Needless to say.

Further, if it is not necessary to stick to the standard technology, any method can be applied as the above-mentioned orthogonal transform or entropy coding method, and the concept of the present invention can be applied to any other coding method. Can also be incorporated. In particular, in the case of an encoding (compression) method including quantization, only by adding a function of adjusting or switching a quantization step, the compression ratio can be finely adjusted in accordance with the distance from a focused color in a hue / saturation space. The effect can be adjusted, and the effect is very large as compared with the method in which the compression ratio can be switched only at a limited stage.

The present invention changes the compression ratio based on hue and saturation. In addition to this, for example, a method of changing the compression ratio according to the distance from a main subject on an image, image quality deterioration, and the like. In particular, a method of changing the compression ratio at an edge portion of an image in which the compression is likely to occur, or a method of changing the compression ratio by a combination of these criteria can be considered. As a method based on the combination, for example, a priority or weight is set to the above-mentioned criterion. For example, first, if the skin color is a skin color based on hue and saturation, the compression ratio is lowered. A method of determining a compression ratio is conceivable.

As described above, with the image compression method and apparatus of the present invention, highly efficient image compression utilizing human visual characteristics can be performed. Further, since the image compression method and apparatus of the present invention can coexist with the standardization technology, it can be said that the image compression method and apparatus are very convenient for users of images.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an outline of an embodiment of the present invention.

FIG. 2 is a block diagram illustrating preprocessing in the embodiment of FIG. 1;

FIG. 3 shows an example of a preprocessing table.

FIG. 4 is a block diagram showing a conventional image encoding / decoding procedure.

[Explanation of symbols]

 1 Preprocessing

Claims (14)

[Claims] 1. An image compression method for compressing image information by encoding image information representing an image, wherein a hue and a saturation of each of a plurality of image regions constituting the image are obtained, Encoding the image information included in the image information at a compression rate determined based on the hue and saturation. 2. The method according to claim 1, wherein the encoding is performed to quantize the image information, and the compression ratio is determined by determining a value of a quantization step of the quantization based on the hue and saturation. 2. The image compression method according to claim 1, wherein: 3. The encoding includes dividing the image information into a plurality of image blocks, performing an orthogonal transform on the image information for each of the image blocks, and applying a transform coefficient obtained by the orthogonal transform to the image information. 3. The image compression method according to claim 2, wherein the quantization is performed by performing the quantization. 4. The image compression method according to claim 3, wherein said hue and saturation are obtained for each of said image blocks. 5. The image compression method according to claim 2, wherein the encoding is performed by performing entropy encoding on the quantized image information. 6. The quantization does not depend on the hue and saturation after performing a pre-processing for determining the compression ratio based on a function or a table determined based on the hue and saturation. 6. The image compression method according to claim 2, wherein predetermined quantization is performed. 7. The compression ratio is reduced when the hue and saturation indicate flesh color.
The image compression method according to claim 1. 8. An image compression apparatus for compressing image information by encoding the image information representing the image, wherein the hue / saturation obtaining means for obtaining hue and saturation for each of a plurality of image regions constituting the image. And an encoding unit that encodes the image information included in each of the image regions at a compression rate determined based on the hue and saturation. 9. The encoding means has a quantization means for quantizing the image information, and the quantization means obtains a value of a quantization step of the quantization by the hue / saturation acquisition means. 9. The image compression apparatus according to claim 8, wherein the compression ratio is determined based on the determined hue and saturation. 10. The encoding device according to claim 1, further comprising: a blocking unit configured to divide the image information into a plurality of image blocks; and an orthogonal transformation unit configured to perform an orthogonal transformation on the image information for each of the image blocks. 10. The image compression apparatus according to claim 9, wherein said quantization means is means for performing said quantization on a transform coefficient obtained by said orthogonal transformation. 11. An image compression apparatus according to claim 10, wherein said hue and saturation are obtained for each of said image blocks. 12. The apparatus according to claim 9, wherein said encoding means further comprises entropy encoding means for further performing entropy encoding on said quantized image information. Image compression device. 13. The pre-processing unit, wherein the quantization unit performs pre-processing for determining the compression ratio based on a function or a table determined based on the hue and saturation, and 13. The image compression apparatus according to claim 9, further comprising a quantizer that performs predetermined quantization independent of the hue and the saturation on the image information. 14. The image compression method according to claim 8, wherein said encoding means is means for reducing said compression ratio when said hue and saturation indicate flesh color. apparatus.