JP3450357B2 - Image processing apparatus and method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image recording apparatus and an image processing apparatus for digitally recording image data of a plurality of screens.

[0002]

2. Description of the Related Art Conventionally, a technique for compressing image data of a plurality of screens and recording it on a recording medium has been known.

As an example of such a technique, a compression method called JPEG method for known image data is known.

When compressing a plurality of image data given by using such a compression method to a fixed length, if the allocation of the code amount per image is L, from the start of continuous shooting to the end of continuous shooting of n images. The code amount L _{1 to} L _n of each image is suppressed to the code amount L or less by repeatedly performing the 2-pass compression method.

[0005]

However, in the above-mentioned conventional example, the data compression work takes a long time because the two-step encoding is performed for each photographing at the time of continuous recording, which is a limitation when increasing the continuous recording speed. .

In view of the above point, in the present invention, it is intended to prevent the speed of continuous recording from decreasing by suppressing the total code amount from the start to the end of continuous recording to a predetermined code amount at high speed during continuous recording. To do.

[0007]

In order to achieve the above object, an image processing apparatus according to the present invention comprises an input means for inputting a series of image data of n (n is a natural number of 4 or more) screens, and the input means. A quantizing means for quantizing the image data input by the means, a coding means for variable-length coding the image data quantized by the quantizing means, and a quantizing step used by the quantizing means. A control unit for controlling each screen, wherein the control unit is pre-assigned to the image data for the n screens a quantization step for the image data of the screens finally encoded in the n screens. N-1 already encoded from the code amount
Control based on the code amount obtained by subtracting the sum of the code amount of the image data for the screen, the quantization step for the image data of the screen other than the screen to be finally encoded, the image of the previous screen of the encoding target screen It is characterized in that the control is performed based on the code amount when the data is encoded.

In order to achieve the above-mentioned object, the image processing method according to the present invention has a series of n (n is a natural number of 4 or more).
An input step of inputting image data of a screen, a quantization step of quantizing the input image data, an encoding step of variable-length encoding the quantized image data, and a use in the quantization step And a control step of controlling the quantization step for each screen, wherein in the control step, the quantization step for the image data of the screen to be finally encoded in the n screens is performed by the image data for the n screens. Already encoded from the code amount previously assigned to
The quantization step for the image data of the screen other than the screen to be finally encoded is controlled based on the code amount obtained by subtracting the sum of the code amounts of the image data for one screen, and It is characterized in that the control is performed based on the code amount when the image data is encoded.

[0009]

FIG. 3 is a block diagram of an electronic still camera according to an embodiment of the present invention, in which 101 is a lens, 102 is a diaphragm, 103 is a shutter, 104 is a solid-state image sensor for converting an image into an electric signal, and 105. Is the output of the solid-state image sensor
An AD conversion circuit for D conversion, and 106 is a memory for temporarily storing data for processing the AD converted signal. Reference numeral 107 denotes a signal processing circuit that creates a luminance signal and a color difference signal by calculation from the output of the solid-state image sensor 104 read from the memory 106 and stores the calculation result in the memory 106 again. 108 indicates the signal from the memory 106 by 8 × 8
Is a DCT conversion circuit for dividing each block into DCT coefficients (discrete cosine transform) into 8 × 8 DCT coefficients and storing in the memory 106. 109 is a DCT for compressing the code amount of the DCT coefficient read from the memory 106.
This is a quantization circuit that quantizes the coefficients. Reference numeral 110 is a quantization table for setting the quantization coefficient used for the quantization of the DCT coefficient. Reference numeral 111 denotes a quantization step adjustment circuit for adjusting the quantization step by multiplying each coefficient of the quantization table. 112 is the quantized D
It is a zigzag scanning circuit that performs zigzag scanning of CT coefficients for each block. Reference numeral 113 is a DPC for further compressing the data amount by taking the difference between the blocks of the DC component of the DCT coefficient.
It is an M circuit. Reference numeral 114 denotes a Huffman coding circuit that performs Huffman coding on the output of the DPCM circuit. A Huffman table 115 is referred to when Huffman coding is performed. 1
Reference numeral 16 is a run length encoding circuit which counts the interval between the non-zero coefficient and the non-zero coefficient in the AC component of the zigzag-scanned DCT coefficient, that is, the length in which zero continues. 117
Is a truncation circuit that forcibly sets high-order coefficients to zero when the code length assigned to each block is about to be exceeded. A Huffman coding circuit 118 assigns a Huffman code to a zero run length and a non-zero coefficient following zero. A Huffman table 119 is referred to by the Huffman coding circuit. A code amount detection circuit 120 detects the code amount. Reference numeral 121 is a memory card for recording compression-encoded data. Based on the code amount detected by the code amount detection circuit 120, the quantization step adjustment circuit 111 determines the coefficient value to be multiplied by each coefficient of the quantization table and the code amount to be assigned to each block in order to obtain the target code amount. . When the code amount of each block is likely to exceed the specified amount, the coding is terminated by forcibly setting the high-order AC coefficient to zero. A system control circuit 122 controls the operation of the entire system. Reference numeral 123 is a release switch for starting shooting, and 124 is a mode switching switch for switching between continuous shooting and single shooting mode.

FIG. 4 is a diagram showing an imaging sequence for shooting one image. When the release 123 is turned on at time T0, the photometric operation is performed from time T0 to T1,
The appropriate shutter speed and aperture value are determined. Next, the shutter 103 is opened at time T1 to T2 to perform pre-exposure, the shutter 103 is closed at time T2, and then the exposure charge is read out. Based on the information on the actual exposure charge between times T3 and T4. The optimum exposure amount is calculated. Next, at time T4, the shutter 103 is opened again to perform main exposure.
Exposure charge reading AD conversion circuit 1 between times T5 and T6
In step 05, AD conversion is performed and the result is stored in the memory 106. Between T6 and T7, the signal is read from the memory 106 to the signal processing circuit 107 to generate a luminance signal (Y), a color difference signal (RY), and a color difference signal (BY). Next, DCT is executed between time T7 and T8.
The conversion circuit 108 performs DCT conversion. From time T8 to T9, compression encoding is performed by a method described later. The compressed image signal is output from the memory card 121 between T9 and T10.
To record.

Prior to the DCT calculation performed at the time of compression, image data is usually divided into blocks of 8 × 8 pixels. FIG. 8 is a diagram showing how the image is divided into blocks. As shown in FIG. 8A, the image is divided into blocks of 8 × 8 pixels in order from the upper left of the screen, and each pixel in each block is indexed by s ₀₀ to s ₇₇ as shown in FIG. 8B. It Figure 9
FIG. 6 is a diagram showing DCT coefficients for each block. By applying the DCT operation of the equation (1) to the signal of 8 × 8 pixels in FIG. 8B, 8 × 8 D is obtained as shown in FIG.
The CT coefficients S _{00 to} S ₇₇ are obtained. Inverse DCT operation is (2)
It is shown by the formula.

[0012]

[Outer 1]

For the coefficients Q _{00 to} Q ₇₇ shown in the quantization table shown in FIG. 9B, the quantization step correction coefficient F
D corresponding to each index Q _'00 ~Q' ₇₇ multiplied by the
The quantized DCT coefficients Sq _{00 to} Sq ₇₇ are obtained by dividing the CT coefficient.

FIG. 10 is a diagram for explaining a method of coding the quantized DCT coefficient. Quantized DC
The T coefficients Sq _{00 to} Sq ₇₇ are zigzag scanned in the order shown in FIG. By doing so, the DCT coefficients are arranged in the order of increasing spatial frequency from Sq ₀₀ indicating the DC component of the block. When the zigzag scanning is performed in this way, Sq ₀₀ indicating a DC component, that is, a DC coefficient and AC coefficients Sq ₀₁ to Sq ₇₇ indicating an AC component are arranged as shown in FIG. 10B. As a general characteristic of an image, an energy component having a high spatial frequency becomes small, and the above-described quantization often causes an AC coefficient to become zero in a high frequency component. Therefore, the data amount of the AC coefficient can be compressed by Huffman coding the pair of the zero run length from the non-zero coefficient to the non-zero coefficient of the zigzag scanned AC coefficient and the non-zero coefficient following the zero. On the other hand, regarding the DC coefficient, the DPCM that takes the difference from the DC coefficient of the adjacent block
The amount of data is compressed by Huffman coding the predicted value. At this time, the size of the code amount depends on the method of quantization. If the quantization is roughly performed, the AC coefficient has many zero components and the amount of data is small, but the image quality is deteriorated.

FIG. 5 is a diagram showing a proportional sequence for showing the effect of the present invention when performing compression coding when the switch 124 is set to the continuous shooting mode. DCT
The encoding using is basically variable length. However, it is general to devise some fixed length so that the number of recordings on the card is constant. Conventionally, it has been customary to use a method called a two-pass method. The two-pass algorithm is shown in FIGS. 6 and 7. FIG. 6 shows the first step of the two-pass method, and FIG. 7 shows the second step of the two-pass method.

A method of fixing length will be briefly described with reference to FIGS. 6 and 7. It is performed before encoding until the image data of each block is converted into DCT coefficients by DCT calculation and stored in the memory. In the first step, the quantization step correction coefficient F is first compared with the quantization table.
The quantization width is set by temporarily setting. Next quantization,
Zigzag scanning and encoding are performed. Next, in order to achieve the target code amount for each block, the code amount of each block is calculated, and the quantization step correction coefficient F in the second step for making the code amount of the entire image the target code amount is predicted. The first step is completed by allocating the maximum code amount for each block. Next, in the second step, the quantization step is set by the quantization step correction coefficient F set in the first step, quantization is performed, zigzag scanning and encoding are performed, but when encoding the AC coefficient, When the allocation of each block determined in the first step is about to be exceeded, the coding amount of the higher-frequency component of the block is cut off so that the code amount does not exceed the set code amount. To convert. Various marker codes necessary for decoding are added to the encoded data, and the second step is completed.

In the above-described method of contrast, the data compression is performed after the two-step encoding, so that the time required for the data compression becomes long.

Next, the coding sequence during continuous shooting in one embodiment of the present invention will be described with reference to FIG. FIG. 2 is a flowchart of encoding by the one-pass method.

The operation of the embodiment of the present invention will be described below with reference to FIGS. 1 and 2. When the release switch 123 is turned on while the mode switching switch 124 is set to the continuous shooting mode, continuous shooting is started, but the first image is encoded by the 1-pass method. The one-pass encoding is performed by the procedure shown in FIG. That is, the quantization step correction coefficient F is set to a predetermined value in advance (S101), quantization (S103), zigzag scanning (S105), encoding (S107 to S113), and marker code addition (S).
119) and the code amount (S115) is calculated. The operation in the case of continuous shooting will be described with reference to FIG. At this time, there is a high possibility that the compressed code amount of the first image does not match the assigned code amount of one image, so the image of the first image is not recorded on the card, and the code of the image assigned per image is not recorded. When the amount is L, the image data before compression is temporarily stored in the memory 106 until the end of the continuous shooting, with an empty space of L left on the card. However, it is assumed that the memory 106 has a capacity of two images. Also, the quantization width of the next image is determined in the process of compressing the first image. During continuous shooting, the pattern does not change significantly between the two subsequent images, so the image to be recorded first should be evaluated to determine the quantization step for compression of the next image to be almost appropriate. You can If the code amount of the image assigned to each image is L, it is highly possible that the first image for continuous shooting does not match L, but the code amount of the second and subsequent images can be brought close to L. . When the release switch 213 is turned off, continuous shooting ends, but the number of continuous shots is n, the assigned code amount per image is L, the code amount of the second image is L ₂ , the code of the nth image is If the amount is L _n , two passes are performed so that L ₁ is a code amount obtained by subtracting the sum of the code amounts from L ₂ to L _n from the code amount that is the product nL of the assigned code amount per continuous shooting number It is compression-encoded according to the method.

FIG. 11 is a diagram showing an encoding sequence during continuous shooting in another embodiment of the present invention.

The operation of the present invention will be described with reference to FIGS. 11 and 2. When the release switch 123 is turned on with the mode switching switch 124 set to the continuous shooting mode, continuous shooting is started, but the first image is encoded by the one-pass method of FIG. 2 described above. In the present embodiment, as shown in FIG. 11, when the code allocation amount per sheet is L, it is 1
The quantization step correction coefficient F is set so that the code amount of the first image becomes a value sufficiently smaller than 2L, and even if the compression code amount of the first image does not match the assigned code amount of one image, it remains as it is. Record. Also, the quantization width of the next image is determined in the process of the first sheet. During continuous shooting, the pattern does not change significantly between the two subsequent images, so the image to be recorded first should be evaluated to determine the quantization step for the compression of the next image to be an appropriate value. You can If the code amount of the image assigned to each image is L, it is highly likely that the first image at the start of continuous shooting does not match L, but the code amount of the second and subsequent images can approach L. Become. When the release switch 213 is turned off, continuous shooting ends, but the number of continuous shots is n, the assigned code amount per image is L, the code amount of the first image is L ₁ , n−.
The code amount of the first image is L _n -1, and the code amount of the _n -th image is L _n -1.
_{If n} , then L _n is set by the two-pass method such that the code amount obtained by subtracting the sum of the code amounts from L ₁ to L _n −1 from the code amount of product nL of the assigned code amount per continuous shot number Compress and encode.

Since the compression of the last image before the end of continuous shooting is performed after the end of continuous shooting, it may take some time. Therefore, the iterative method may be such that the code amount becomes L _n by repeating the 1-pass system encoding while changing the quantization step correction coefficient F until the code amount converges to L _n , instead of the 2-pass system.

The present invention can be applied not only to the electronic still camera but also to an image file device, a color copying machine, etc., and is not limited to the electronic still camera.

[0024]

As described above, according to the present invention, high-speed encoding processing can be performed while controlling the code amount of a series of image data for n (n is a natural number of 4 or more) screens to a predetermined amount. In addition, the code amount pre-allocated to the image data for n screens can be effectively used without waste.

[Brief description of drawings]

FIG. 1 is a diagram showing an encoding sequence during continuous shooting according to the present invention.

FIG. 2 is a flowchart of encoding by the one-pass method.

FIG. 3 is a block diagram of an electronic still camera.

FIG. 4 is a diagram showing an image capturing sequence for capturing one image.

FIG. 5 is a diagram showing a comparative sequence of the present invention when performing compression encoding when the switch 124 is set to the continuous shooting mode.

FIG. 6 is a diagram showing a first step of a two-pass method.

FIG. 7 is a diagram showing a second step of the two-pass method.

FIG. 8 is a diagram showing how an image is divided into blocks.

FIG. 9 is a diagram showing DCT coefficients for each block.

FIG. 10 is a diagram illustrating a method of encoding a quantized DCT coefficient.

FIG. 11 is a view showing the sequence of another embodiment of the present invention.

[Explanation of symbols]

108 DCT conversion circuit 109 Quantization circuit 112 Zigzag scanning circuit

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04N ⁷ /24-7/68 H04N 1/41-1/419 H04N 5/91-5/956 H04N 5 / 78-5/781 H03M 7/30-7/50

Claims (2)

(57) [Claims] 1. An input means for inputting a series of image data of n (n is a natural number of 4 or more) screens, a quantizing means for quantizing the image data input by the input means, and the quantizing means. The encoding means for variable-length encoding the quantized image data, and the control means for controlling the quantization step used by the quantization means in screen units, the control means in the n screen The quantization step for the image data of the screen finally encoded is
The control is performed based on the code amount obtained by subtracting the sum of the code amounts of the already coded image data of n-1 screens from the code amount assigned in advance to the image data of the screen, and finally encoding is performed. An image processing apparatus, wherein a quantization step for image data of a screen other than the screen to be controlled is controlled based on a code amount when the image data of the previous screen of the encoding target screen is encoded. 2. An input step of inputting a series of image data of n (n is a natural number of 4 or more) screens, a quantization step of quantizing the input image data, and a step of quantizing the quantized image data. A coding process for variable-length coding and a control process for controlling the quantization step used in the quantization process on a screen-by-screen basis. In the control process, a screen to be finally coded in the n screens. To the code amount obtained by subtracting the sum of the code amounts of the already coded image data of n-1 screens from the code amount previously assigned to the image data of n screens. Based on the code amount when the image data of the previous screen of the encoding target screen is encoded, the quantization step for the image data of the screen other than the screen to be finally encoded. Image processing method for the butterflies.