JP3458412B2 - Quantization bit number conversion device and conversion method, and quantization bit number learning device and learning method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a quantizing bit number converting apparatus suitable for use in, for example, a digital video tape recorder (digital VTR) or the like, and more particularly to a quantizing bit number for an externally supplied image signal. Number signal conversion device and method for converting the number of image signals into a plurality of image signals for output , and quantization bit mathematics
A learning device and a learning method .

[0002]

2. Description of the Related Art Today, 8 bits are the most common quantization bit number in the digitization of image signals. For example, the CCIR / Rec 601 which is an encoding standard for component television signals also specifies that the number of quantization bits is 8 bits. This was decided based on various visual experiments in consideration of factors such as image quality, hardware scale, and storage medium consumption.

However, in the case of increasing the demand for higher image quality and performing special effect processing such as image synthesis, transformation and enlargement / reduction, image quality deterioration due to insufficient gradation of 8-bit quantization and various signal processing in the camera. From the viewpoint of ensuring the accuracy of calculation at the time, it is desired to shift the number of quantization bits from 8 bits to 10 bits. One example is SMPTE.2, which is a standard for serial digital interfaces.
In 59M, 10 bits have already become the standard.

At present, the image recording apparatus is mainly used for processing and recording 8-bit data, and it is difficult to expand to 10 bits due to the problem of the amount of recording medium consumed. For example, in a digital video recorder such as the D1 format, when a 10-bit signal is input during recording, only the upper 8-bit data is recorded. Then, at the time of reproduction, the quantization bit number conversion device simply adds the code of 00 as the lower 2 bits of the 8-bit data and converts the 8-bit data to 8-bit.

[0005]

However, in the conventional quantization bit number conversion device, the data length of each data is simply 8 bits.
Since the number of bits is simply increased to 10 bits, the output signal has no difference in the amount of information from the data coded with 8 bits from the beginning, and there is a problem such as occurrence of pseudo contour due to insufficient gradation.

Therefore, the object of the present invention is made in view of the above-mentioned problems, and not only the number of quantization bits is increased, but also the signal of a small number of quantization bits is dealt with to increase the information amount. Quantized bit number conversion device and conversion method capable of converting into
To provide a learning method and a learning method .

[0007]

According to a first aspect of the present invention, a pattern of a level distribution of an image signal is detected for each block of input image data, and a class to which the image signal belongs based on the detected pattern. And class conversion means for outputting class information and conversion information for converting into image data quantized with more quantization bits than the input image data are stored for each class. The input image data is converted into image data having a larger number of quantization bits based on the storage unit for outputting the conversion information and the conversion information output from the storage unit according to the class information from the determining unit. It is a quantization bit number conversion device characterized by comprising a conversion means. According to a fifth aspect of the present invention, the conversion input image data is converted into the conversion input image.
Image with more quantization bits than data quantization bits
Quantization, which learns conversion information for converting to image data
A bit number learning device, which is the amount of input image data for conversion
Bed for blocking takes out pixels from input image data composed of many quantization bits than Coca bits
The locking means and the pattern of the level distribution of the image signal are detected for each block of the blocked image data , the class to which the image signal belongs is determined based on the detected level distribution pattern, and the class information is obtained. Class determining means for outputting, counting means for counting the lower bit data pattern of the input image data for each class information, lower bit data which is the maximum from the lower bit data patterns counted for each class information And a storage unit that stores the detected lower bit data pattern for each class information. According to a ninth aspect of the present invention, the pattern of the level distribution of the image signal is detected for each block of the input image data, the class to which the image signal belongs is determined based on the detected pattern, and class information is output. , Conversion information for converting to image data quantized with a quantization bit number larger than the input image data is stored for each class, and the conversion information is output according to the class information from the class determination means. A method for converting a quantization bit number is characterized in that input image data is converted into image data having a larger number of quantization bits based on the output conversion information. In the invention according to claim 13, the conversion input image data is converted into an input image for conversion.
Image with more quantization bits than data quantization bits
Quantization, which learns conversion information for converting to image data
A bit number learning method, which is the amount of input image data for conversion
Into blocks from the input image data consisting of many quantization bits than Coca number of bits extracts predetermined pixels, blanking
The pattern of the level distribution of the image signal is detected for each block of the locked image data, the class to which the image signal belongs is determined based on the detected level distribution pattern, and class information is output. The lower-order bit data pattern of the data is counted for each class information, and the maximum lower-order bit data pattern is detected from the lower-order bit data patterns counted for each class information, and the detected lower-order bit data pattern is detected. Is stored for each class information, which is a quantized bit number learning method.

[0008]

The quantized bit number conversion device according to the present invention converts the image data quantized with a larger number of bits than the input image data by using the lower bit data obtained by learning for each image signal pattern. By doing so, problems such as the occurrence of false contours due to insufficient gradation can be solved.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the quantization bit number converter according to the present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram of a learning unit of the quantization bit number conversion device according to the present invention. Learning data corresponding to a known image quantized with 10 bits is supplied to the input terminal 1. In this case, a plurality of images with different designs are used. The image data input from the input terminal 1 is processed by the blocking circuit 2 to extract pixels having an arrangement relationship as shown in FIG. In this example, one block is a (3 × 3) array centered on the pixel of interest (processing target pixel). Figure 4 is centered on the target pixel x ₀ of 10-bit (3 × 3) surrounding 8 pixels x ₁ ~x included in the block
₈ shows the situation of a block of 9 pixels in total. When the processing is completed for one target pixel, the target pixel is shifted by one pixel and the block is also shifted by one pixel as a whole.

The lower bit separation circuit 3 stores the lower 2 bits of the 10-bit image data supplied from the blocking circuit 2.
The upper 8 bits obtained by separating the bits are supplied to the class code generation circuit 4, and the lower 2 bits are supplied to the counter 5.

As will be described later, the class code generating circuit 4 outputs the comparison result regarding the level change in the block and the class code c, respectively. Class code generation circuit 4
The comparison result output from is supplied to the counter 5. The counter 5 measures the number of appearances of the bit pattern (00, 01, 10, 11) of the lower 2 bits in a predetermined period. However, when the comparison result of the pixels supplied from the class code generation circuit 4 becomes the threshold Th or more, the counter 5 does not count the number of appearances of the bit pattern of the lower 2 bits.

The count values from the counter 5 are supplied to the frequency memory 6, respectively. The frequency memory 6 stores the count value supplied from the counter 5 with the class code c supplied from the class code generation circuit 4 as an address address. The maximum value detection circuit 7 detects the maximum value from the count values stored in the frequency memory 6. The lower bit generation circuit 8 generates a 2-bit pattern corresponding to the maximum count value. In addition, the data memory 9
Then, the class code c supplied from the class code generating circuit 4 is supplied as an address address and the bit pattern supplied from the lower bit generating circuit 8 is stored.

An example of the class code generation circuit 4 shown in FIG. 1 will be described with reference to FIG. The image data of upper 8 bits out of 10 bits supplied from the input terminal 11 is supplied to the detection circuit 12 and the subtractor 14, and the detection circuit 1
In 2, the maximum value and the minimum value in the above (3 × 3) block are detected, respectively. The detected maximum value is supplied to the subtractor 13, and the detected minimum value is supplied to the subtractor 13 and the subtractor 14, respectively. The subtractor 13 subtracts the minimum value from the maximum value detected by the detection circuit 12, and the output thereof is supplied to the comparison circuit 15. That is, each of the supplied data of pixels x _{0 to} x ₈ (8 bits) of one block is set to e ₀ to e ₈ , and the maximum value and the minimum value thereof are respectively set to e.
_{Let max} and e _min . At this time

Considering the case of e _max −e _min <4 (1), e _i ′ = e _i −e _min (i = 0 ,.
.. 8) is data that can be represented by 2 bits.

The comparison circuit 15 compares the supplied data with the threshold value Th input from the terminal 16 and outputs the comparison result from the output terminal 17. In this example, the threshold value Th is 4. The subtractor 14 subtracts the minimum value detected by the detection circuit 12 from the input data, and the conversion circuit 1
8 and outputs the class code c from the output terminal 19. The formula of the process for calculating the class code c performed by the conversion circuit 18 is shown below.

[0016]

[Equation 1]

Here, p = 2 (bits).

The learning section is not limited to the hardware configuration shown in FIG. 1, but may have a software processing configuration as shown in the flowchart of FIG. The control of the learning unit is started from step 21, and in the data block formation of step 22, a process of extracting a plurality of pixels of one block having the above-mentioned arrangement relationship is performed. At the end of the data in step 23, if learning of all the input data, for example, one frame of data has been completed, the lower bit data determination of step 26 is performed, and if there is more data to be learned, the class determination of step 24 is performed. Control is transferred to.

In the class determination in step 24, the learning data is divided into classes according to the signal pattern.
The simplest method for this class division is to use the upper 8-bit sequence of 10-bit data in a block as the class number as it is. However, this method requires provision of a huge capacity of ROM. Therefore, the quantization bit number conversion using the lower bit data for each class is performed. In particular, the learning process is limited to a portion in which the level change is small in which a problem such as image quality deterioration is likely to occur when lower bits are reduced. That is, even if the conversion from 8 bits to 10 bits is performed by simply inserting the code of 00 in the lower 2 bits in the portion where the level fluctuation is large, the deterioration is hardly noticeable due to the masking effect.

In the lower bit count of step 25,
The number of appearances of each bit pattern of the lower 2 bits in a predetermined period is counted. After the processing of all data is completed, the control moves from step 23 to step 26. Step 26
In the determination of the lower bit data of, the bit pattern of the lower 2 bits corresponding to the bit pattern having the highest number of appearances in the lower bit count in step 25 is determined as the lower bit and held. At this time, step 24
The bit pattern is held by using the class code c determined by the class determination of as the address address.

FIG. 5 is a block diagram of an embodiment of the quantization bit number converter according to the present invention. In FIG. 5, 3
An 8-bit digital image signal quantized into 256 gradations is supplied to the input terminal indicated by 1. This input signal is sent by transmission by broadcasting or reproduction by a video tape recorder device or the like. The digital image signal from the input terminal 31 is supplied to the blocking circuit 32. The block formation circuit 32 converts the input data in the order of the raster scan of the television into the data in the order of the block as shown in FIG. The 8-bit output data of the blocking circuit 32 is the detection circuit 33, the 00 addition circuit 34,
It is supplied to the lower bit addition circuit 35 and the class code generation circuit 36, respectively.

The detection circuit 33 detects the maximum value and the minimum value from the supplied blocked data, and the subtractor 37 subtracts the minimum value from the maximum value output from the detection circuit 33.
Output to the comparison circuit 38. In the comparison circuit 38, the supplied data is input to the terminal 39 through a threshold value Th (=
4), and the comparison result is supplied to the selector 40 as a control signal. This subtractor 37 and the comparison circuit 3
The process performed by 8 is the same as that represented by the above-mentioned formula (1).

The 00 addition circuit 34 adds 00 as the lower 2 bits of the supplied block data,
That is, 10-bit data is output to the selector 40.

The class code generation circuit 36 determines the class to which the image data of the block belongs based on the pattern of the two-dimensional level distribution of the image of each block supplied by performing the operation of the equation (2). A class code c which is the detected class detection information is generated, and the class code c is output to the data memory 9 in which the above learning result is stored. The data memory 9 outputs the bit pattern of the lower 2 bits stored in the address with the class code c as the address, and outputs it to the lower bit addition circuit 35. In the lower bit addition circuit 35, the bit pattern of the lower 2 bits supplied from the data memory 9 is added to the block data of 8 bits supplied from the blocking circuit 32, and the output data thereof, that is, 10 bits. Data is supplied to the selector 40.

Here, the selector 40 includes the 00 addition circuit 3
Data obtained by converting the number of quantization bits from 8 bits to 10 bits are supplied from the 4 and lower bit addition circuits 35, respectively. The selector 40 uses the formula (1) to determine the comparison circuit 38.
Is a threshold Th (threshold Th =
4) If 10 or more, the 10-bit data output from the 00 addition circuit 34 is selectively output to the output terminal 41. If the comparison result is less than 4, the 10-bit data output from the lower bit addition circuit 35 is selectively output to the output terminal 41. Output.

FIG. 6 is a flow chart showing the operation of the quantization bit number conversion device according to the present invention when it is configured as software processing. The control of the conversion unit is started from step 51, and in the data block formation of step 52, the process of extracting one block of pixels having the above-mentioned arrangement relationship is performed. In step 53, if the processing of all the input data, for example, one frame of data has been completed, the process proceeds to the end of step 56, and if not completed, the control proceeds to the class determination of step 54.

In the class determination in step 54, the class is determined from 8-bit data. In the lower bit addition in step 55, the class determined in step 54 is used as an address address, the bit pattern stored in the address address is read, and the read bit pattern is added to the 8-bit data.

[0028]

The quantized bit number converter according to the present invention is information for converting image data supplied from the outside into image data quantized with a quantized bit number larger than that of the image data. The lower bit data is stored in the lower bit data storage means. Then, the pattern of the two-dimensional level distribution is detected for each block of the 8-bit image data supplied from the outside, the lower bit data corresponding to the pattern of the image information is read from the lower bit data storage means, and the external The image data and the lower bit data supplied from are added and output. The present invention can not only increase the number of quantization bits, but can also convert an increase in the number of quantization bits into a signal corresponding to an increase in the amount of information, without the occurrence of pseudo contours due to insufficient gradation.

[Brief description of drawings]

FIG. 1 is a block diagram of an example of a configuration of a learning unit in an image signal conversion device according to the present invention.

FIG. 2 is an example of a block diagram used for explaining a class code generation circuit in the image signal conversion apparatus according to the embodiment of the present invention.

FIG. 3 is a flowchart of an example of a configuration of a learning unit in the image signal conversion device according to the present invention.

FIG. 4 is an example of a schematic diagram used for explaining image data according to an embodiment of the present invention.

FIG. 5 is a block diagram of an example of a configuration of an image signal conversion device according to the present invention.

FIG. 6 is a flowchart of an example of the configuration of the image signal conversion device according to the present invention.

[Explanation of symbols]

9 data memory 32 block circuit 33 Detection circuit 3400 additional circuit 35 Lower bit addition circuit 36 class code generator 38 Comparison circuit 40 selector

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T 5/00 100 G06T 7/00 H04N 1/40

Claims (16)

(57) [Claims] 1. A class determination for detecting a level distribution pattern of an image signal for each block of input image data, determining a class to which the image signal belongs based on the detected pattern, and outputting class information. Means, conversion information for converting to image data quantized with a quantization bit number larger than the input image data is stored for each class, depending on the class information from the class determining means, Storage means for outputting the conversion information, based on the conversion information output from the storage means,
A quantized bit number conversion device, comprising: a conversion unit for converting the input image data into image data having a larger number of quantized bits. 2. The quantized bit number conversion apparatus according to claim 1, wherein said storage means stores lower bit data, and said conversion means adds said lower bit data. 3. When the level change of the input image data is smaller than a predetermined threshold value, the conversion means reads the conversion information from the storage means, adds the conversion information, and outputs the input image data. 2. The quantized bit number conversion apparatus according to claim 1, wherein when the level change of 1 is above the threshold value, 0 is simply added as the conversion information. 4. The quantized bit number converter according to claim 1, wherein said class determining means detects a two-dimensional level distribution of said image signal. 5. The input image data for conversion is input for the conversion.
Quantization bit number more than the quantization bit number of image data
The amount of learning conversion information for converting to image data
A Coca bits learning device, pixels from the input image data consisting of many quantization bits than the number of quantization bits of the converted input image data
A block forming means for taking out the image data into blocks, and detecting a pattern of the level distribution of the image signal for each block of the block of the image data , and the image signal belongs based on the detected pattern of the level distribution. Class determining means for determining a class and outputting class information, counting means for counting the pattern of the lower-order bit data of the input image data for each of the class information, and lower-order bits for each of the class information It is characterized by comprising a detection means for detecting the pattern of the lower-order bit data that is the maximum among the data patterns, and a storage means for storing the detected pattern of the lower-order bit data for each of the class information. Quantization bit number learning device. 6. The quantized bit number learning apparatus according to claim 5, further comprising a lower bit separating means for separating the lower bit data from the input image data. 7. The counting means is adapted not to count the pattern of the lower bit data when the level change of the input image data is equal to or more than a predetermined threshold value. Quantization bit number learning device described. 8. The quantized bit number learning device according to claim 5, wherein said class determining means detects a two-dimensional level distribution of the image signal. 9. A pattern of a level distribution of an image signal is detected for each block of input image data, a class to which the image signal belongs is determined based on the detected pattern, and class information is output. Conversion information for converting to image data quantized with a quantization bit number larger than that of data is stored for each class, and the conversion information is output according to the class information from the class determining unit. A method for converting the number of quantized bits, wherein the input image data is converted into image data having a larger number of quantized bits based on the output conversion information. 10. The quantization bit number conversion method according to claim 9, wherein the conversion information is lower bit data, and the lower bit data is added to the input image data. 11. When the level change of the input image data is smaller than a predetermined threshold value, the stored conversion information is read out, the conversion information is added, and the level change of the input image data is set as described above. 10. The method for converting the number of quantized bits according to claim 9, wherein 0 is simply added as the conversion information when the threshold value or more. 12. The method according to claim 9, wherein a two-dimensional level distribution of the image signal is detected when the class is determined. 13. Inputting the conversion input image data to the conversion input image data.
More quantized bits than the number of quantized bits in the force image data
Learning conversion information for converting into a number of image data,
A quantization bit number learning method, pixels from the input image data consisting of many quantization bits than the number of quantization bits of the converted input image data
To obtain a block, detect the pattern of the level distribution of the image signal for each block of the blocked image data , and determine the class to which the image signal belongs based on the detected level distribution pattern. The class information is output, the pattern of the lower-order bit data of the input image data is counted for each of the class information, and the maximum of the lower-order bit data of the lower-order bit data patterns counted for each of the class information is calculated. A quantized bit number learning method, wherein a pattern is detected, and the detected pattern of the lower bit data is stored for each class information. 14. The quantization bit number learning method according to claim 13, further comprising separating the lower bit data from the input image data. 15. The quantized bit according to claim 13, wherein when the level change of the input image data is equal to or more than a predetermined threshold value, the pattern of the lower bit data is not counted. Number learning method. 16. The method for learning a quantization bit number according to claim 13, wherein when the class is determined, a two-dimensional level distribution of an image signal is detected.