JP3849906B2 - Image compression apparatus and image compression method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image compression apparatus and an image compression method , and more particularly to an image compression apparatus and an image compression method that contribute to saving storage capacity in an image input apparatus having a limited storage capacity such as a digital camera.
[0002]
[Prior art]
With the development of video cameras and digital still camera devices that have been exclusively used as video input devices, there has been an increasing demand to use them as document input devices. In particular, recent digital still cameras have begun to be used for various purposes as portable information collection tools due to their high pixels and small size. Character information such as documents, signboards, and advertisements can be obtained as a binary image because sufficient information can be obtained with a binary image and the storage capacity required for storage is small. In addition, the binarized image can be transmitted by fax, or can be reused after character recognition. Further, when a 3 million pixel class CCD is used, even if the multi-valued color image (YUV 4: 2: 2) is reduced to 1/16 by JPEG compression, 375 kilobytes are required.
[0003]
On the other hand, if it is a black and white binary image, it is 375 kilobytes even if it is not compressed, and if binary image compression such as MH and MMR is performed, sufficient compression can be expected if the image is centered on characters.
[0004]
[Problems to be solved by the invention]
However, when an image such as a natural image is binarized using pseudo halftone processing such as dithering or error diffusion, the code length after MH or MMR compression is often larger than the image size before compression, meaning that compression is performed. There is no. In principle, JBIG does not become larger than the image size before compression, but the calculation is complicated and the compression rate does not significantly improve for a pseudo-halftone image.
[0005]
The present invention has been made in view of the above circumstances,
An object of the present invention is to perform a fixed-length multi-value image compression and a variable-length binary image compression for compressing a binary image to a predetermined code length, and compare the code length after compression by both compressions. An object of the present invention is to provide an image compression apparatus and an image compression method for efficiently compressing an image by selecting a code having a shorter code length.
[0006]
Another object of the present invention is to provide an image input apparatus capable of storing as many black and white binary images as possible using the compression apparatus of the present invention.
[0007]
[Means for Solving the Problems]
In the present invention, 7 bits having a value of 0 are added to the lower order of the binary image to convert it to 8 bits (value of 0 or 255), and the 8-bit image is compressed by a fixed length compressor. Further, the binary image is compressed by the variable length compressor, and it is checked whether or not the code length exceeds a predetermined byte. When the predetermined byte is exceeded, the code compressed by the fixed length compressor is selected, and when the predetermined byte is not exceeded, the code compressed by the variable length compressor is selected.
[0008]
In addition, when the present invention is applied to an image input device, after binarizing the captured multi-valued image, the encoded length compressed by the fixed length compressor is compared with the encoded length compressed by the variable length compressor. Then, the shorter code is selected and stored.
[0009]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.
Example 1
FIG. 1 shows the configuration of Embodiment 1 of the present invention. In the figure, 1 is a bit converter, 2 is a fixed length compressor, 4 is a variable length compressor, 3, 5 is a buffer, 6 is a selector, and 7 is a code length discrimination signal.
[0010]
The binary image is input to the variable length compressor 4. The variable length compressor 4 performs compression such as MH or MMR, which is a known binary image compression method. MH and MMR have the advantage that they are simple and easy to implement, and the compression speed is fast. In particular, since the compression method is optimal for character images, a high compression rate can be obtained for document images. For example, in an image having a size of 1280 × 960, the code size of MH and MMR is about 20 kbytes to 30 kbytes while the image size of the uncompressed binary image is 154 kbytes. In the case of a binary image subjected to pseudo halftone processing, the code length after compression is larger in both cases of MH and MMR than the image size before compression.
[0011]
The binary image is converted to 8 bits by adding 7 bits having a value of 0 to the lower order in the bit converter 1. The value of the image converted to 8 bits is either 0 or 255. The 8-bit image is compressed by the fixed length compressor 2. The fixed length compressor 2 performs known JPEG compression. JPEG compression is originally a multi-value image compression method for 8-bit color images, but if binary values 0 and 1 are assigned to luminance values 0 and 255 and the color component is set to 0, the application to the binary image is Is possible. The code length after JPEG compression is set to about 2/3 of the image size of the uncompressed binary image. For example, if the image size is 1280 × 960, the image size of the uncompressed binary image is 154 kbytes. If the image compression ratio is set to 2/3, the code length after JPEG compression is 100 kbytes. If it is about 100 kbytes, even if the JPEG image is decoded and binarized again, the deterioration is not conspicuous compared to the binary image before encoding.
[0012]
The codes encoded by the variable length compressor 4 and the fixed length compressor 2 are temporarily stored in the buffers 5 and 3, respectively. The variable length compressor 4 outputs a determination signal 7 to the selector 6 as to whether or not the code exceeds 100 kbytes. If the selector 6 receives the discrimination signal 7 from the variable length compressor 4 and exceeds 100 kbytes (that is, the code length of the variable length encoding is longer than the code length of the fixed length encoding), the selector 6 If the code does not exceed 100 kbytes (that is, the code length of variable length coding is shorter than the code length of fixed length coding), the code of the buffer 5 is read and output.
[0013]
(Example 2)
FIG. 2 shows a configuration of the second embodiment of the present invention. FIG. 2 is a diagram showing a configuration from image input of a digital still camera to binarized compression storage of luminance signals.
[0014]
The CCD 11 is means for converting the light collected through the optical system into an electrical signal, and outputs an analog image signal of each pixel R, G, B. The AGC circuit 10 is a circuit that electrically amplifies the signal output from the CCD, and in the case of shooting where the amount of exposure is insufficient, such as when the subject is insufficiently bright and the shutter speed cannot be slowed down, the gain is increased. Ensure that a moderate level of output is obtained. The output image signal is converted into a digital signal by the A / D converter 12. The white balance of the digital signal is adjusted by the white balance adjuster 13. White balance adjustment is to detect an achromatic color portion in an image and adjust each color signal so that the achromatic color becomes R = G = B.
[0015]
Next, a luminance signal is generated by a luminance signal generation unit that combines the luminance signal generator 14 and the luminance signal selector 15. The luminance signal generator 14 generates two luminance signals for the image whose white balance has been adjusted. One is a wideband luminance signal, and the other is a narrowband luminance signal. The broadband luminance signal YH is calculated by the following equation when the target pixel is R, G, B, respectively.
YH_R = R
YH_G = G
YH_B = B
When the subject is black and white, the broadest luminance signal is generated.
[0016]
Next, for the narrowband luminance signal YL, for example, when RGB filters are arranged in a checkered pattern as shown in FIG. 3, the YL signal in G0 is generated as shown in the following equation.
YL_G0 = (4G0 + G1 + G2 + G3 + G4) / 8
The YL signal at R0 and B0 is generated as in the following equation.
YL_R0 = (G0 + G1 + G2 + G5) / 4
YL_B0 = (G0 + G1 + G3 + G6) / 4
[0017]
As described above, the surrounding G pixels are used to generate a narrow-band luminance signal YL with less noise by taking an average. The luminance signal selector 15 has an Ev value calculated from the aperture amount and shutter speed of the camera for obtaining proper exposure and a preset value Ev0 (the f value of the lens when the aperture is open and a slow shutter speed that does not cause camera shake). When the Ev value is greater than Ev0, YH is selected, and when the Ev value is less than Ev0, YL is selected and output.
[0018]
However, in actual shooting, in order to prevent camera shake without using a flash, the shutter speed is set to 1 / focal length (mm) seconds at the latest.
[0019]
The generated luminance signal corrects the contour portion of the image by the aperture corrector 16. FIG. 4 shows a detailed configuration of the aperture corrector. For the luminance signal, vertical and horizontal Laplacian filters 31 and 32 calculate a vertical Laplacian and a horizontal Laplacian. The absolute values 33 and 34 of the Laplacian in the vertical direction and the horizontal direction are compared by the comparator 35, the Laplacian having the larger absolute value is selected by the selector 36, and added to the original luminance signal by the adder 37.
[0020]
As shown in FIG. 5, the coefficient of the Laplacian filter is the high-pass type shown in FIG. 5A when the Ev value as the exposure amount is larger than a predetermined value Ev0, and the band-pass type filter shown in FIG. Use. The filter coefficient is stored in the parameter 38, and the filter coefficient (a) or (b) is set in the filters 31 and 32 according to the comparison result. If the band-pass type is used, fine noise of one dot is suppressed, and when the brightness of the subject is insufficient, not only background noise is reduced but also the quality of characters is improved.
[0021]
Next, the brightness-corrected luminance signal is binarized. The aperture corrected luminance signal is stored in the temporary frame memory 17. Since the image size is determined by the size of the CCD 11, based on the image size, the CPU 18 calculates a block size and a sampling period necessary for the binarization process, and a block buffer 19 and a non-low brightness pixel average value calculator 20 respectively. To control.
[0022]
FIG. 6 is a diagram showing the sampling period. FIG. 6A shows a sampling period when the image size is small, and FIG. 6B shows a sampling period when the image size is large. The unevenness of the brightness of the image taken by the camera is bright near the center of the image and becomes darker as it goes to the periphery. Therefore, if the shape of the block division of the image is stored in advance in the ROM of the CPU 18 so as to divide the center of the image as a symmetric point with a combination of square, rectangle, and triangle as shown in FIG. It becomes closer to uniformity and allows higher quality binarization compared to using only square blocks.
[0023]
The block buffer 19 is a buffer that reads an image from the frame memory and stores it temporarily in units of blocks having a preset shape and size. The non-low-brightness pixel average value calculator 20 is a unit that samples and reads out pixels from the image stored in the block buffer 19 at a preset sampling period and calculates the average of the luminance values.
[0024]
FIG. 8 shows a detailed configuration of the non-low luminance pixel average value calculator. The sampled pixel value is compared with the threshold value thl_ij by the comparator 41. If the pixel value is larger than the threshold value as a result of the comparison, the gate 45 is opened, the value of the addition result register 42 and the pixel value are added by the adder 46, and the result is stored in the addition result register 42. At the same time, the counter 43 is incremented by one. Here, when the counter 43 is in a state where the digit is incremented as a result of the increment (a state in which the counter indicates a power of 2), the gate 47 is opened, the result of the adder is stored in the shift register 44, and the bit indicated by the counter Shift right by a number. After processing all the pixels in the block, the value stored in the shift register 44 is output as the non-low luminance pixel average value ave_ij.
[0025]
The non-low luminance pixel detection threshold value setting unit 21 multiplies the non-low luminance pixel average value ave_i-1j of the previous block (indicating (i-1, j) if the current block is (i, j)) by a predetermined coefficient Ca. Then, the threshold value thl_ij used by the non-low luminance pixel average value calculator 20 is calculated. If the predetermined coefficient is Ca = 1/4, the non-low luminance pixel detection threshold value setting unit 21 is a value excluding the lower 2 bits of ave_i-1j, so that no special circuit is required. Using the ave_ij calculated by the non-low luminance pixel average value calculation circuit 20, the binarization threshold setting circuit 22 sets a binarization threshold th_ij for image binarization.
[0026]
The binarization threshold setting circuit 22 is a multiplier that multiplies a predetermined coefficient Cb by ave_ij, and Cb = x / 16 or Cb = x / 8 (x is a default value that represents a natural number not exceeding the denominator). For example, since the binarization threshold value setting circuit 22 can be composed of only an adder, it is advantageous in terms of cost and speed. The binarizer 23 binarizes the block using the binarization threshold th_ij. The binarized image is compressed by the compressor 24.
[0027]
The detailed structure of the compressor is shown in FIG. The binarized image is input to the variable length compressor 51 and the fixed length compressor 52. The variable length compressor 51 performs compression such as MH or MMR, which is a known binary image compression method. MH and MMR have the advantage that they are simple and easy to implement, and the compression speed is fast. In particular, since the compression method is specialized for character images, a high compression ratio can be obtained for document images. For example, in an image having a size of 1280 × 960, the code size of MH and MMR is about 20 kbytes to 30 kbytes while the image size of the uncompressed binary image is 154 kbytes. In the case of a binary image subjected to pseudo halftone processing, both the MH and MMR have a larger code length after compression than the image size before compression.
[0028]
The fixed length compressor 52 performs known JPEG compression. JPEG compression is originally a multi-value image compression method for 8-bit color images. However, if binary values 0 and 1 of a binary image are assigned to luminance values 0 and 255 and the color component is set to 0, application to a binary image is possible. Is possible. The code length after JPEG compression is set to about 2/3 of the image size of the uncompressed binary image. For example, if the image size is 1280 × 960, the image size of the uncompressed binary image is 154 kbytes. If the image compression ratio is set to 2/3, the code length after JPEG compression is 100 kbytes. If it is about 100 kbytes, even if the JPEG image is decoded and binarized once again, it is at a level that does not bother degradation compared to the binary image before encoding.
[0029]
The codes encoded by the variable length compressor 51 and the fixed length compressor 52 are temporarily stored in the buffers 53 and 54, respectively. The variable length compressor 51 outputs a determination signal 56 to the code selector 55 as to whether or not the compressed code exceeds 100 kbytes. If the code selector 55 receives the discrimination signal 56 of the variable length compressor 51 and exceeds 100 kbytes (if the code length of the variable length encoding is longer than the code length of the fixed length encoding), the code selector 55 If the code does not exceed 100 kbytes (when the code length of variable-length coding is shorter than the code length of fixed-length coding), the code of the buffer 53 is read and output to the memory 25 of FIG. The saved image is saved.
[0030]
Example 3
FIG. 10 shows the configuration of the third embodiment of the present invention. The difference from the second embodiment is that a pseudo halftone processing binarization circuit 27, a selector 28, and a mode setting unit 9 are provided. The process until the aperture corrected luminance signal is stored in the temporary frame memory 17 is the same as in the second embodiment.
[0031]
The aperture-corrected luminance signal is binarized as follows. As in the second embodiment, the image is transferred from the frame memory 17 to the block buffer 19. The image transferred to the block buffer 19 is input to the pseudo halftone processing binarization circuit 27 and the simple binarization circuit 23. The pseudo halftone processing binarization circuit 27 binarizes the image using pseudo halftone processing (known dither processing or error diffusion processing).
[0032]
The block buffer 19, the non-low luminance pixel average value calculator 20, the non-low luminance pixel detection threshold value setting unit 21, and the binarization threshold value setting circuit 22 in the character binarization circuit 26 are the same as those described in the second embodiment. Therefore, explanation is omitted.
[0033]
Similar to the second embodiment, the block is binarized by the simple binarizer 23 using the binarization threshold th_ij. The pseudo halftone processed binary image and the simple binarized binary image are input to the selector 28.
[0034]
The mode setting unit 9 sets the character binarization mode or the natural image binarization mode of the camera. When the character binarization mode is set, a simple binarized binary image is When the natural image binarization mode is set, a binary image subjected to pseudo halftone processing is selected by the selector 28 and output. The selected image is compressed by the compressor 24 in the same manner as described in the second embodiment, and the encoded image is stored in the memory 25.
[0035]
Example 4
FIG. 11 shows the configuration of the fourth embodiment of the present invention. The difference from the third embodiment is that an image area separation circuit 30 is provided instead of the mode setting device.
[0036]
As in the second embodiment, the image transferred to the block buffer 19 is input to the image area separation circuit 30, the pseudo halftone processing binarization circuit 27, and the simple binarization circuit 23. The image area separation circuit 30 determines a character or non-character (design) area for each pixel of the image.
[0037]
FIG. 12 shows details of the image area separation circuit 30. An image signal from the block buffer 19 and an output signal from the simple binarization circuit 23 are input to the image area separation circuit 30. The input image signal is output by an edge extraction circuit 301, and a Laplacian operation (the pixel of interest is a central pixel) having a mask coefficient as shown in FIG. 13 or a first-order differential operation as shown in FIG. The binarization circuit 302 binarizes the calculated value edge with a predetermined threshold th to determine whether the pixel of interest is an edge or a non-edge.
[0038]
In addition, the boundary between the white and black of the output signal of the simple binarization circuit 23 is detected by the boundary detection circuit 303. The boundary detection circuit 303 determines that the target pixel is a boundary when both the white pixel and the black pixel are mixed in the 3 × 3 pixel including the target pixel, and determines that the white pixel is included in the 3 × 3 pixel. When only a pixel or only a black pixel exists, it is determined that it is not a boundary.
[0039]
By taking the logical product 304 of the output of the binarization circuit 302 and the output of the boundary detection circuit 303, it is determined whether or not the pixel of interest is a character region (edge and boundary) (non-character region). Output to.
[0040]
Returning to FIG. 11, the pseudo halftone processing circuit 27 binarizes the image using pseudo halftone processing (known dither processing or error diffusion processing).
[0041]
The binary image subjected to pseudo halftone processing and the simple binarized binary image are based on the character / non-character discrimination signal of the image area separation circuit 30, and in the case of the character area, the simple binarized binary image is obtained. In the case of a non-character area, a binary image subjected to pseudo halftone processing is selected by the selector 28 and output. The selected image is compressed by the compressor 24 in the same manner as described in the second embodiment, and the encoded image is stored in the memory 25.
[0042]
(Example 5)
The fifth embodiment of the present invention is an embodiment related to an image compression method. FIG. 15 is a process flowchart of the fifth embodiment. First, a binary image is input (step 101). The input binary image is converted into an 8-bit multi-value image that takes only binary values of 0 and 255 (7-bit 0 is added to the lower order), and fixed for a multi-value image at a predetermined compression rate. Long coding is performed (step 102). A typical example of fixed length coding is JPEG.
[0043]
Next, variable length coding for the binary image is performed on the binary image (step 103). Typical examples of variable length coding are MH and MMR. The code lengths encoded in the respective encodings are compared (step 104). If the code length of variable length coding is short, a variable length code is selected (step 105), and if the code length of fixed length coding is short, a fixed length code is selected (step 106) and stored (step 107).
[0044]
【The invention's effect】
As described above, according to the present invention , a fixed-length multi-value image compression having a simple configuration and a high compression rate can be performed on a binary image.
[0045]
According to the present invention , since different compression methods or compression means are prepared, it is possible to obtain image compression with a high compression rate and high image quality regardless of the contents of the binary image.
[0046]
According to the present invention , it is possible to obtain an image input device capable of storing a binarized image with high compression and high image quality when storing a captured image.
[0047]
According to the present invention , when binarizing an image, the binarizing means can be switched according to the type of image to be input, so that a high-quality binary image can be obtained.
[0048]
According to the present invention , since a binarization method can be selected for each area according to the area (character / non-character) of the image, a high-quality binary image can be obtained.
[Brief description of the drawings]
FIG. 1 shows a configuration of Embodiment 1 of the present invention.
FIG. 2 shows a configuration of Embodiment 2 of the present invention.
FIG. 3 shows a pixel arrangement of a CCD.
FIG. 4 shows a configuration of an aperture correction circuit.
FIG. 5A shows a high-pass Laplacian, and FIG. 5B shows a band-pass Laplacian.
FIGS. 6A and 6B show different sampling periods, respectively.
FIG. 7 shows an image block dividing method.
FIG. 8 shows a configuration of a non-low luminance pixel average value calculator.
FIG. 9 shows a configuration of an image compressor.
FIG. 10 shows a configuration of Embodiment 3 of the present invention.
FIG. 11 shows a configuration of Example 4 of the present invention.
FIG. 12 shows a configuration of an image area separation circuit.
FIG. 13 shows Laplacian mask coefficients.
FIG. 14 shows a first derivative operation.
FIG. 15 shows a processing flowchart of Embodiment 5 of the present invention.
[Explanation of symbols]
1 bit converter 2 fixed length compressor 3, 5 buffer 4 variable length compressor 6 selector 7 code length discrimination signal

Claims (4)

A block dividing unit that divides input multi-valued image data into a plurality of blocks, and a binarization threshold value is determined for each block divided by the block dividing unit according to the brightness of pixels in the block. A simple binarization unit that binarizes pixel data in a block by a binarization threshold; a pseudo halftone binarization unit that converts each pixel data of the input multivalued image into a pseudo halftone binary image; and the input An image area separating unit that determines whether each pixel of the multi-valued image is a character area or a non-character area, and the pixel that is determined to be a character area by the image area separating unit The pixel data binarized by the binarizing means is selected, and the pixel data binarized by the pseudo halftone binarizing means is selected for the pixels determined to be non-character areas by the image area separating means. A first selection means to select; Variable length compression means for performing variable length compression on binary image data composed of pixel data selected by the selection means, and binary image data composed of pixel data selected by the selection means Multi-value conversion means for converting to multi-value image data having a predetermined bit width; fixed-length compression means for performing fixed-length compression on the multi-value image data obtained by the multi-value conversion means; and the variable length compression And an image compression apparatus comprising: a second selection unit that compares a code length after compression by the fixed length compression unit and selects a code having a shorter code length. A block dividing unit that divides input multi-valued image data into a plurality of blocks, and a binarization threshold value is determined for each block divided by the block dividing unit according to the brightness of pixels in the block. A simple binarization unit that binarizes pixel data in a block by a threshold value, a pseudo halftone binarization unit that converts each pixel data of the input multivalued image into a pseudo halftone binary image, Binarization mode setting means for setting either the binarization mode or the natural image binarization mode; and when the character binarization mode is set by the binarization mode setting means, the simple binarization When the pixel data binarized by the means is selected and the natural image binarization mode is set by the binarization mode setting means, the pixel binarized by the pseudo halftone binarization means First choice to select data A variable-length compression unit that performs variable-length compression on binary image data composed of pixel data selected by the selection unit, and a binary image composed of pixel data selected by the selection unit Multi-value conversion means for converting data into multi-value image data having a predetermined bit width; fixed-length compression means for performing fixed-length compression on multi-value image data obtained by the multi-value conversion means; and the variable An image compression apparatus comprising: a long compression unit; and a second selection unit that compares a code length after compression by the fixed length compression unit and selects a code having a shorter code length. A block dividing step for dividing the input multi-valued image data into a plurality of blocks, and for each block divided by the block dividing step, a binarization threshold value is determined according to the brightness of the pixels in the block, and the binary A simple binarization step for binarizing pixel data in a block by a threshold value, a pseudo halftone binarization step for converting each pixel data of the input multi-valued image into a pseudo halftone binary image, and the input An image area separating step for determining whether each pixel of the multi-valued image is a character area or a non-character area, and for the pixels determined to be a character area in the image area separating step, the simple binary The pixel data binarized by the binarization step is selected, and the pixel data binarized by the pseudo halftone binarization step is selected for the pixel determined to be a non-character region by the image area separation step. A first selection step to select; and A variable length compression step for performing variable length compression on binary image data composed of pixel data selected by the selection step, and binary image data composed of pixel data selected by the selection step A multi-value conversion step for converting to multi-value image data having a predetermined bit width, a fixed-length compression step for performing fixed-length compression on the multi-value image data obtained by the multi-value conversion step, and the variable length compression And a second selection step of comparing a code length after compression in the fixed length compression step and selecting a code having a shorter code length. A block dividing step for dividing the input multi-valued image data into a plurality of blocks, and the brightness of the pixels in the block for each block divided by the block dividing step. A simple binarization step of binarizing the pixel data in the block with the binarization threshold, and converting each pixel data of the input multi-valued image into a pseudo halftone binary image A pseudo halftone binarization step for conversion, a binarization mode setting step for setting either a character binarization mode or a natural image binarization mode, and a character binarization mode in the binarization mode setting step Is set, the pixel data binarized by the simple binarization step is selected, and when the natural image binarization mode is set by the binarization mode setting step, Variable length compression is performed on a first selection step for selecting pixel data binarized by the halftone binarization step and binary image data composed of the pixel data selected by the selection step. Variable length compression process and selected by the selection process Multi-value conversion process for converting binary image data composed of pixel data into multi-value image data having a predetermined bit width, and fixed-length compression for multi-value image data obtained by the multi-value conversion process A second selection step of comparing a code length after compression by the variable length compression step and the fixed length compression step and selecting a code having a shorter code length. An image compression method characterized by the above.

Images