JP2017037353A - Image processing device and control method thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently execute local image processing.SOLUTION: An image processing apparatus divides an input image into plural band region images, and successively stores each band region image in a band memory to perform image processing. The image processing apparatus includes: storage means which stores a first filter coefficient having a reference region of M pixels in the first direction in the input image and one pixel in the second direction different from the first direction, and a second filter coefficient having a reference region of one pixel in the first direction and N pixels (N<M) in the second direction; division means which divides the band region images stored in the band memory into K divided images by extracting an image group for each K line at the mutually-different phases; spatial filter means which sequentially applies the primary filter using the first filter coefficient and the primary filter using the second filter coefficient to each of K divided images; and generation means which generates an image after the filter processing to the band region images by compositing the K processed divided images to which the filter processing is applied by the spatial filter means.SELECTED DRAWING: Figure 6

Description

  The present invention relates to an image processing technique using a spatial filter.

  Conventionally, local image processing such as spatial filter processing has been performed when an image is formed and output. This local image processing is image processing in which some calculation is performed using all pixels included in a reference region including a pixel to be processed (hereinafter referred to as a processing pixel). For example, spatial filter processing such as edge enhancement processing and blur processing is performed on the digital image data.

  In such local image processing, processing is performed from the left end to the right end for the uppermost pixel column of the digital image data, and then processing is performed for the second pixel row from the upper end, and the same operation is repeated until the lowermost pixel column. There are cases where a method of performing is used. In this method, a larger memory capacity is required as the width of the digital image data in the main scanning direction increases. For example, when an A4 size image is read by a scanner having a resolution of 600 dpi and converted to digital image data, the width of the digital image data is 4953 pixels. Therefore, when a pixel has a data amount of 3 bytes (24 bits) and local (neighboring) image processing having a reference area of 3 pixels × 3 pixels is performed, about 29 Kbytes (4953 pixels × 2 lines × 3) Bytes) of memory capacity.

  Therefore, in order to reduce the memory capacity required for image processing, there is a technique (hereinafter referred to as block processing) in which digital image data is divided into a plurality of block (tile) areas and separately (locally) image processing is performed. (Patent Document 1). Further, in Patent Document 2, a method of performing image processing by scanning pixels in each band area with respect to a plurality of band areas obtained by dividing one piece of image data in the sub-scanning direction (hereinafter referred to as cross-band). Called processing). Furthermore, in Patent Document 3, in order to reduce the amount of computation of two-dimensional spatial filter processing, a method of separating the two-dimensional spatial filter processing into one-dimensional filters each in the main scanning direction and the sub-scanning direction (hereinafter, variable separation). Called type filter processing).

JP 11-259646 A Japanese Patent No. 45944022 JP 2011-97258 A

  However, in the methods described in Patent Documents 1 and 2, since there are regions (overlapping regions) that overlap the boundaries of the regions, there is a problem that the total number of transfer pixels increases if the reference region in the local image processing is widened. . For example, when recovering an image in which optical system blur has occurred due to spatial filtering, sufficient recovery performance can be obtained unless the reference area is expanded to the same extent as the optical system blur, as shown in FIG. Absent. Therefore, when the blur of the optical system is large, it is necessary to widen the reference area, and as a result, the total number of transfer pixels increases. Alternatively, when it is desired to remove the noise of the imaging system by blurring the image by spatial filter processing, it is necessary to widen the range to be blurred as the noise increases. Therefore, when the noise of the imaging system is large, it is necessary to widen the reference area, and as a result, the total number of transfer pixels increases. At that time, even if the variable separation filter processing described in Patent Document 3 is used, the overlapping region does not change. Therefore, the increase amount of the total transfer pixel number does not change.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide a technique that can execute local image processing more efficiently.

In order to solve the above-described problems, an image processing apparatus according to the present invention has the following configuration. That is, in an image processing apparatus that divides an input image into a plurality of band area images and sequentially stores each band area image in a band memory to perform image processing.
A first filter coefficient having a reference region of M pixels in the first direction and a single pixel in a second direction different from the first direction in the input image, one pixel in the first direction, and the second Storage means for storing a second filter coefficient having a reference area of N pixels (where N <M) in the direction of
Dividing means for dividing the band area image stored in the band memory into K divided images by extracting pixel groups for each K rows at different phases;
Spatial filter means for sequentially applying a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K divided images;
Generating means for synthesizing K post-processed divided images to which the filter processing by the spatial filter means is applied, and generating a post-filtering image for the band region image;
Have

Alternatively, in an image processing apparatus that divides an input image into a plurality of band area images and sequentially stores each band area image in a band memory to perform image processing.
A first filter coefficient having a reference region of M pixels in the first direction and a single pixel in a second direction different from the first direction in the input image, one pixel in the first direction, and the second Storage means for storing a second filter coefficient having a reference region of M pixels in the direction of
Dividing means for dividing the band area image stored in the band memory into K divided images by extracting pixel groups for each K rows at different phases;
Spatial filter means for sequentially applying a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K divided images;
Generating means for synthesizing K post-processed divided images to which the filter processing by the spatial filter means is applied, and generating a post-filtering image for the band region image;
Have
At least one of the first direction and the second direction is a direction having an inclination with respect to the arrangement direction of the pixel groups in the row direction in the input image.

  ADVANTAGE OF THE INVENTION According to this invention, the technique which enables local image processing to be performed more efficiently can be provided.

1 is a block diagram illustrating an overall configuration of an image processing apparatus according to a first embodiment. It is a block diagram which shows the structure of an image process part. It is a figure explaining the band area | region in a band process. It is a figure explaining the operation | movement of the band process by a variable separation type filter. It is a figure explaining the form which reduces the resolution of one direction of image data to 1/2. It is a figure explaining operation | movement of the filter process in 1st Embodiment. It is a figure which shows the structure of the one-dimensional filter in 1st Embodiment. It is a figure which shows the area | region where a spatial filter process is performed exemplarily. It is a figure explaining the characteristic after correction | amendment by the conventional filter process. It is a figure explaining the characteristic after correction | amendment by the filter process which concerns on 1st Embodiment. It is a figure explaining the form which reduces the resolution to 1/4 in one of two one-dimensional filters. It is a figure explaining the form which reduces the resolution to 1/2 in both of two one-dimensional filters. It is a figure which shows the structure of the one-dimensional filter in 3rd Embodiment. It is a figure explaining the relationship between the blur range by optical blur, and the influence range of the filter coefficient of a spatial filter. 3 is a diagram illustrating a functional configuration of a spatial filter circuit 230. FIG.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. The following embodiments are merely examples, and are not intended to limit the scope of the present invention.

(First embodiment)
A first embodiment of an image processing apparatus according to the present invention will be described below by taking as an example a multifunction peripheral (MFP) that reads a document, corrects (compensates) image blur caused by the reading optical system, and forms an image. .

<Prerequisite technology>
First, the image processing method described in the background art which is the premise of the present invention will be described.

Block processing This is a technique in which digital image data is divided into a plurality of block (tile) areas, and local (neighboring) image processing is performed separately. In such a technique, in order to perform local (neighboring) image processing without any gap between the block areas, the block areas overlap with each other adjacent block areas at the boundary. In this way, the capacity of the memory can be defined by the size of the block (tile) area, and the memory can be saved. For example, it is A4 size digital image data (4969 pixels in the main scanning direction × 7016 pixels in the sub-scanning direction) with a resolution of 600 dpi. At this time, when performing spatial filter processing having a reference area of 3 pixels × 3 pixels, assuming that the size of the block (tile) area is 16 pixels × 16 pixels, the memory capacity is 0.75 Kbytes (16 pixels × 16 Pixel × 3 bytes).

  However, the pixels in the overlapping area are transferred twice or four times. For example, when the spatial filter process is performed on the reference area of 3 pixels × 3 pixels, the total number of transfer pixels is 1.31 times the original number of pixels of digital image data, 4969 pixels × 7016 pixels. In addition, when performing spatial filter processing on a 9 × 9 spatial filter area, the number of overlapping areas further increases, and the total number of transfer pixels of digital image data is four times the number of pixels of the original digital image data. It becomes.

-Cross-band processing A plurality of band areas obtained by dividing one image data in the sub-scanning direction are scanned with pixels in the band area in a direction perpendicular to the main scanning direction of the assigned band area. In this method, local (neighboring) image processing is sequentially performed. With this technique, the memory capacity for image processing of pixels in the band region can be made to depend on the size of the band region (the height of the band region) in the direction perpendicular to the main scanning direction of the band region. . That is, in order to perform local (neighboring) image processing without any gap between the band regions, the band regions overlap each other at the boundary with the adjacent band regions. Unlike the block processing described above, the overlapping areas are not meshed. Therefore, the number of retransfers of the same pixel is small compared to the block processing, and as a result, the total number of transfer pixels is also reduced.

Variable separation type filter processing The two-dimensional spatial filter processing is separated into one-dimensional filters each in the main scanning direction and the sub-scanning direction, first the one-dimensional filter processing in the main scanning direction is performed, and then one-dimensional in the sub-scanning direction. This is a technique for performing filter processing. Since the memory capacity for the reference area may be one column in the main scanning direction or the sub-scanning direction, it can be greatly reduced. However, if sharpness recovery processing using a variable separation filter is performed in the vertical and horizontal directions, the oblique direction is overcorrected (diagonal ringing occurs) because it is different from the two-dimensional optical blur characteristic. Further, although the memory capacity and the calculation amount can be suppressed, the total number of transfer pixels cannot be reduced.

<Outline of First Embodiment>
In the first embodiment, variable separation filter processing is performed on image data (odd row image and even row image) thinned out in the sub-scanning direction, and one of the two one-dimensional filters (sub-scanning direction) is a reference pixel. To reduce. With this configuration, it is possible to reduce overcorrection in the oblique direction. In addition, since the number of pixels used for the spatial filter processing is reduced, it is possible to suppress an increase in the total number of transfer pixels.

<Device configuration>
FIG. 1 is a block diagram showing the overall configuration of the image processing apparatus according to the first embodiment. The image reading unit 120 includes a lens 122, a CCD sensor 124, an analog signal processing unit 126, and the like. An image of the original 100 formed on the CCD sensor 124 via the lens 122 is converted into analog electrical signals of R (Red), G (Green), and B (Blue) by the CCD sensor 124.

  The image information converted into the analog signal is input to the analog signal processing unit 126, and is subjected to analog-digital conversion (A / D conversion) after correction or the like is performed for each color of R, G, and B. A digitized full color signal (hereinafter referred to as a digital image signal) is input to the image processing unit 130. The image processing unit 130 performs an input correction process, a spatial filter process, a color space conversion, a density correction process, and a halftone process, which will be described later, on the digital image signal, and outputs the digital image signal after these processes are performed. The data is output to the printer unit 140. The printer unit 140 includes, for example, a print output unit (not shown) such as a raster plotter that uses an inkjet head, a thermal head, or the like, and records an image on paper using an input digital image signal.

  The CPU circuit unit 110 includes a CPU 112 for arithmetic control, a ROM 114 for storing fixed data and programs, a RAM 116 used for temporarily storing data and loading programs, an external storage device 118, and the like. Then, the image reading unit 120, the image processing unit 130, the printer unit 140, and the like are controlled to comprehensively control the sequence of the image processing apparatus. The external storage device 118 is a medium such as a disk for storing parameters and programs used by the image processing apparatus, and the data and programs in the RAM 116 may be loaded from the external storage device 118.

  FIG. 2 is a block diagram illustrating a configuration of the image processing unit. The digital image signal output from the analog signal processing unit 126 is input to the image processing controller 200 via the bus 205. The image processing controller 200 includes an input interface 210, an input correction circuit 220, a spatial filter circuit 230, a color space conversion circuit 240, a density correction circuit 250, a halftone processing circuit 260, and an output interface 270. Hereinafter, the input correction circuit 220, the spatial filter circuit 230, the color space conversion circuit 240, the density correction circuit 250, and the halftone processing circuit 260 will be described in detail.

  A digital image signal 215 is input to the input correction circuit 220 via the input interface 210. The digital image signal 215 is composed of R, G, and B luminance signals. The input correction circuit 220 performs processing for correcting variations in characteristics of the sensor that reads the document 100 and light distribution characteristics of the document illumination lamp.

  The digital image signal (luminance signals R, G, B) 225 output from the input correction circuit 220 is input to the spatial filter circuit 230. The spatial filter circuit 230 is a circuit corresponding to the main characteristic portion of the first embodiment, and the digital image signal (luminance signals R, G, B) 225 is locally (such as smoothing, edge enhancement, optical blur correction). Neighboring) Image processing is performed.

  A digital image signal (luminance signals R, G, B) 235 output from the spatial filter processing circuit 230 is input to the color space conversion circuit 240. The color space conversion processing circuit 240 converts the luminance signals R, G, and B of the digital image signal 235 into density signals C (Cyan), M (Magenta), Y (Yellow), and K (Black).

  A digital image signal (density signals C, M, Y, K) 245 output from the color space conversion circuit 240 is input to the density correction circuit 250. The density correction circuit 250 performs density correction on the digital image signal (density signals C, M, Y, K) 245. This is because it is necessary to perform density correction in advance in consideration of the characteristics of halftone processing so that density change does not occur when binarization is performed by the subsequent halftone processing circuit 260.

  A digital image signal (density signals C, M, Y, K) 255 output from the density correction circuit 250 is input to the halftone processing circuit 260. The halftone processing circuit 260 performs screen processing on the digital image signal (density signals C, M, Y, K) 255 and converts it into a binary halftone expression. A binary digital image signal (print signals C, M, Y, K) 265 is output to the printer unit 140 via the output interface 270 and the bus 275.

<Operation of Spatial Filter Circuit 230>
In a low-cost device such as a home printer, the capacity of the system main memory (corresponding to the RAM 116 in FIG. 1) is small. Therefore, it is common to divide one piece of image data into a plurality of band (band-like) areas and develop each band area in a main memory to perform various image processes. This divided and elongated area is called a band area, a storage area where the data in the band area is expanded is called a band memory, and the act of dividing is called band division. The band memory is not necessarily reserved as a storage area in the main memory, and may be allocated in any storage area on the system. However, for the sake of brevity, the band memory is used as the main memory. An example of securing in the memory will be described.

  FIG. 3 is a diagram for explaining the operation of the filter process in the band process. Here, the coordinate system (main scanning direction-sub-scanning direction) of the digital image data defines a new coordinate system (band area coordinate system) of the length direction and the height direction as shown in FIG. The band region is expressed by length × height. The length of the band area is always a value of the width of the digital image data in the main scanning direction or the height of the sub scanning direction, and the height of the band is an arbitrary value.

  The band processing will be described in a little more detail. First, the first band area 301, which is the first band area image, is developed in the band memory on the main memory, and image processing is performed (FIG. 3A). Next, the second band area 302 is overwritten and expanded on the band memory in which the band area 301 is expanded to perform image processing (FIG. 3B). Further, the third band area 303 is overwritten and expanded on the band memory in which the band area 302 is expanded (FIG. 3C). In the same manner, the image data is sequentially stored and processed up to the lower area of the image data.

  The band regions 301 to 303 have the same length (main scanning direction), but the height (sub-scanning direction) is not necessarily the same. The band memory which is the storage area of the main memory is determined by the largest band area. As described above, the band memory in the main memory is not limited to one storage area. For example, two band memories A and B may be secured in the main memory. In such a case, first, the band area 301 is developed in the first band memory A and the first image processing is performed. Next, the band region 301 is moved from the first band memory A to the second band memory B, and the second band region 302 is developed in the first band memory A. Then, it is preferable to perform the first image processing on the second band region 302 in parallel while performing the second image processing on the first band region 301. By dividing the digital image data into band area units and performing image processing, such pipelined image processing becomes possible.

<Band Processing Operation of Spatial Filter Circuit 230>
In a low-cost device such as a home printer, the capacity of the system main memory (corresponding to the RAM 116 in FIG. 1) is small. Therefore, it is common to divide one piece of image data into a plurality of band (band-like) areas and develop each band area in a main memory to perform various image processes. This divided and elongated area is called a band area, a storage area where the data in the band area is expanded is called a band memory, and the act of dividing is called band division. The band memory is not necessarily reserved as a storage area in the main memory, and may be allocated in any storage area on the system. However, for the sake of brevity, the band memory is used as the main memory. An example of securing in the memory will be described.

  FIG. 15 is a diagram illustrating the functional configuration of the spatial filter circuit 230. The spatial filter circuit 230 includes a band region dividing unit 2301, a pixel group coordinate information determining unit 2302, a filter coefficient holding unit 2303, a product-sum operation unit 2304, and a band region combining unit 2305.

  FIG. 3 is a diagram for explaining the operation of the filter process in the band process. Here, the coordinate system (main scanning direction-sub-scanning direction) of the digital image data defines a new coordinate system (band area coordinate system) of the length direction and the height direction as shown in FIG. The band region is expressed by length × height. The length of the band area is always a value of the width of the digital image data in the main scanning direction or the height of the sub scanning direction, and the height of the band is an arbitrary value.

  The band processing will be described in a little more detail. First, the first band area 301 shown in FIG. 3A is developed in the band memory on the main memory to perform image processing. Next, the second band area 302 shown in FIG. 3B is developed by overwriting the band memory in which the first band area 301 is developed, and image processing is performed. Further, the third band area 303 shown in FIG. 3C is developed by overwriting the band memory in which the second band area 302 is developed, and image processing is performed. Finally, the fourth band area 304 shown in FIG. 3D is overwritten on the band memory in which the third band area 303 is expanded to perform image processing.

  The band regions 301 to 303 have the same length (main scanning direction), but the height (sub-scanning direction) is not necessarily the same. The band memory which is the storage area of the main memory is determined by the largest band area. As described above, the band memory in the main memory is not limited to one storage area. For example, two band memories A and B may be secured in the main memory. In such a case, first, the band area 301 is developed in the first band memory A and the first image processing is performed. Next, the band region 301 is moved from the first band memory A to the second band memory B, and the second band region 302 is developed in the first band memory A. Then, it is preferable to perform the first image processing on the second band region 302 in parallel while performing the second image processing on the first band region 301. By dividing the digital image data into band area units and performing image processing, such pipelined image processing becomes possible.

  FIG. 4 is a diagram for explaining the operation of band processing by a conventional variable separation filter. FIG. 7 is a diagram illustrating a configuration of a one-dimensional filter. Here, an example is shown in which one-dimensional filter (product-sum operation) processing in the main scanning (length) direction is performed, and then one-dimensional filtering processing in the sub-scanning (height) direction is performed. However, the order of processing in the main scanning and sub-scanning directions may be first. In FIG. 4A, reference numeral 400 indicates the entire area of the image data. First, a one-dimensional filter process in the main scanning direction (length direction) is performed using the one-dimensional filter coefficient 410 shown in FIG.

  As shown in FIG. 4A, image processing is executed pixel by pixel along the length direction of the band area from the upper left pixel of the band area. When the pixel reaches the rightmost pixel of the band area, the pixel for image processing is advanced by one pixel in the height direction, and image processing is performed again from the leftmost pixel of the band area to the right edge one pixel along the length of the band area. Executed. This series of image processing is executed up to the lower right pixel of the band area, and the band processing in the main scanning direction is completed. After the processing of the first band region 301 shown in FIG. 4A, the one-dimensional shown in FIG. 7B is applied to the first band region 301 as shown in FIG. 4B. Using the filter coefficient 411, one-dimensional filter processing in the sub-scanning (height direction) is performed.

  As shown in FIG. 4B, image processing is executed pixel by pixel along the height direction of the band area from the pixel at the upper left end of the band area. When reaching the lowermost pixel of the band area, the pixel for image processing is advanced by one pixel in the length direction, and image processing is performed again from the uppermost pixel of the band area to the lower edge one by one along the height direction of the band area. Executed. This series of image processing is executed up to the lower right pixel of the band area, and the band processing in the sub-scanning direction is completed. After finishing the processing for the first band region 301, as shown in FIGS. 4C and 4D, the second band region 302 was performed on the first band region 301. Perform the same process as.

  FIG. 8 is a diagram exemplarily showing a region where the spatial filter process is executed. In the periphery of the digital image data 400, there is an area 1190 that cannot be subjected to spatial filtering. This is because all the pixel values in the reference region by the two one-dimensional filter coefficients shown in FIG. 7A and FIG. 7B cannot be substituted for the spatial filter processing. In principle, appropriate spatial filtering cannot be performed. Therefore, in order to perform local (neighboring) image processing without a gap between the band regions (band regions 301 to 304), overlapping pixel regions 460 as shown in FIG. 4E are essential in each band region. .

  In the example of FIG. 4 described above, an 8-line band memory is required to process 4 lines. That is, in the conventional processing, the digital image data is divided and subjected to local (neighboring) image processing, so that the total number of transfer pixels is larger than the total number of pixels of the digital image data. The increase rate r of the total number of transfer pixels is expressed by Equation (1), as can be seen from FIG. 4E, where the height of the spatial filter region 411 is fh and the band height Bdh which is the height of the band region. Street.

Rate of increase: r = (Bdh) / (Bdh− (fh−1)) (1)
That is, in the conventional processing, since fh = 5 and Bdh = 8, the increase rate r of the total number of transfer pixels is r = 2, and double transfer is required for the actual effective processing area.

<Details of Filter Processing of Spatial Filter Circuit 230>
Below, the filter process of the spatial filter circuit 230 in 1st Embodiment is demonstrated. Here, a case where the height direction of the band memory is the same as the conventional one (FIG. 4) (8 lines) will be described. Specifically, the filtering process is performed on the image thinned out for each row using the spatial filter region 411 in which the reference pixels in the thinned direction are reduced.

  FIG. 5 is a diagram for explaining a mode in which the resolution in one direction of image data is reduced to ½. This processing is executed by the band area dividing unit 2301 and the pixel group coordinate information determining unit 2302. Here, a pixel group is extracted every other row (every two pixels) and divided into two divided images (even-numbered row image 401 and odd-numbered row image 402). However, a group of pixels may be extracted for each K rows, where K is an arbitrary positive integer of 2 or more, and divided into K divided images.

  In FIG. 5A, reference numeral 400 indicates the entire area of the image data. First, the pixel group coordinate information determination unit 2302 assigns even-numbered pixels of the digital image data 400 to the even-numbered image 401 and assigns odd-numbered pixels to the odd-numbered image 402. At this time, the number of lines to be allocated is 16 (= 8 × 2) lines.

  After the assignment of the even-numbered line image 401 and the odd-numbered line image 402 is finished, the band area dividing unit 2301 extracts the even-numbered line image 401 as shown in FIG. To do. Here, the height Bdh of the band memory is “8” as in the conventional case (FIG. 4). The even line image 401 is obtained by thinning out the image data 400 for each line, and the resolution in the height direction is ½ that of the image data 400. A reduction in the total number of transfer pixels is realized by performing a one-dimensional filter process on the even-line image 401. Details will be described later with reference to FIG.

  After completing the one-dimensional filter processing for the even-numbered line image 401, the band area dividing unit 2301 extracts the odd-numbered line image 402 as shown in FIG. 5C and develops it in the band memory as data of the band area 301. . Again, the height Bdh of the band memory is “8”. The odd line image 402 is also obtained by thinning out the image data 400 for each line, and the resolution in the height direction is ½ that of the image data 400. The odd-numbered row image 402 is also reduced by reducing the total number of transfer pixels by performing the one-dimensional filter process in the same manner as the even-numbered row image 401. Details will be described later.

  FIG. 6 is a diagram for explaining the operation of the filter processing in the first embodiment. As described above, in the first embodiment, the filtering process is performed on the data (even-numbered row image 401 and odd-numbered row image 402) obtained by thinning out the image data 400 for each row. Therefore, as the filter coefficient in the height direction (height fh = N pixels), the filter coefficient 412 (FIG. 7C) whose resolution is ½ with respect to the coefficient of FIG. Positive integer). Note that the filter coefficient 410 (FIG. 7A) is used as the filter coefficient in the length direction (length fw = M pixels) as in the conventional case (M is a positive integer). These filter coefficients are held in a storage unit such as the filter coefficient holding unit 2303 in FIG.

  The product-sum operation unit 2304 performs one-dimensional filter processing in the main scanning direction (length direction) on the even-numbered row image 401 stored in the band region 301 using the one-dimensional filter coefficient 410. As shown in FIG. 6A, image processing is performed pixel by pixel along the length direction of the band area from the pixel at the upper left end of the band area. When the pixel reaches the rightmost pixel of the band area, the pixel for image processing is advanced by one pixel in the height direction, and image processing is performed again from the leftmost pixel of the band area to the right edge one pixel along the length direction of the band area. Executed. This series of image processing is executed up to the lower right pixel of the band area, and the band processing in the main scanning direction for the even-numbered row image 401 is completed.

  After finishing the processing shown in FIG. 6A, the product-sum operation unit 2304 performs sub-scanning (height direction) one-dimensional filter processing on the even-numbered image 401 using the one-dimensional filter coefficient 412. . As shown in FIG. 6B, image processing is executed pixel by pixel along the height direction of the band area from the pixel at the upper left end of the band area. When reaching the lower end pixel of the band area, the pixel for image processing is advanced by one pixel in the length direction, and image processing is performed again from the upper end pixel of the band area to the lower end one by one along the height direction of the band area. Executed. This series of image processing is executed up to the lower right pixel of the band area, and the band processing in the sub-scanning direction for the even-numbered row image 401 is completed.

  After finishing the processing for the even-numbered row image 401, the product-sum operation unit 2304 uses the one-dimensional filter coefficient 410 for the odd-numbered row image 402 stored in the band region 301, as shown in FIG. A one-dimensional filter process in the main scanning direction (length direction) is performed. Note that the processing in FIG. 6C is the same as the processing in FIG.

  After completing the processing shown in FIG. 6C, the product-sum operation unit 2304 uses the one-dimensional filter coefficient 412 for the odd row image 402 as shown in FIG. (One direction) one-dimensional filter processing is performed. Note that the process in FIG. 6D is the same as the process in FIG. This series of image processing is executed, and the band processing in the sub-scanning direction for the odd-numbered row image 402 ends.

  For the even-numbered row image 401 and the odd-numbered row image 402 for which the processing shown in FIGS. 6A to 6D has been completed, the band region combining unit 2305 returns the thinned image to the coordinates of the original image 400. That is, two post-processed divided images to which the filter process is applied are synthesized, and a post-filter process image for the first band region 301 is generated. The generated data is input to the color space conversion circuit 240. After the processing for the first band region 301 is completed, the same processing as that performed for the first band region 301 is performed for the second band region 302.

  As described above, the processing target image (even-numbered row image 401, odd-numbered row image 402) and filter coefficient 412 both have a resolution in the height direction of ½. When the resolution in the height direction is ½, the control performance of the high-frequency component in the height direction is deteriorated as compared with the conventional case. However, since the spatial ranges exerted by the filter coefficients in FIG. 7C and FIG. 7A are the same, the control performance of the low frequency component is substantially the same. That is, the filter coefficient 412 shown in FIG. 7C is difficult to control for the high frequency component but easy for the low frequency component. In human visual characteristics, it is known that high-frequency image components are less sensitive than low-frequency image components. Therefore, it can be expected that the filter processing using the filter coefficient 412 shown in FIG. 7C has substantially the same performance as the conventional filter coefficient (FIG. 7B).

<Effect>
Reduction of over-correction in the oblique direction The filter coefficient 412 shown in FIG. 7C can be expected to have a more appropriate filter effect than the conventional filter coefficient shown in FIG. 7B. This can be easily understood by considering a filter (product-sum operation) process for correcting the spatial frequency degradation of the image reading unit 120 as the degradation of the frequency characteristics of the system. In general, the image reading unit 120 includes a lens 122, a CCD sensor 124, and the like. Here, it is known that the image of the original 100 formed on the CCD sensor 124 through the lens 122 is deteriorated in terms of spatial frequency.

  FIG. 9 is a diagram for explaining the characteristic after correction by the filter processing of the conventional variable separation filter. FIG. 10 is a diagram for explaining characteristics after correction by the filter processing according to the first embodiment. FIGS. 9A and 10A are diagrams illustrating an example of the spatial frequency characteristics of the G component of the RGB signal obtained by the CCD sensor 124. FIG. In general, optical blur such as a lens causes isotropic frequency degradation. Therefore, the same frequency degradation occurs in each direction of fx (0 degree direction), fy (90 degree direction), and fxy (45 degree direction).

  Now, let us consider a case where the above-mentioned optical blur is corrected (recovered) with a combination of conventional one-dimensional filters (FIGS. 7A and 7B). The horizontal and vertical one-dimensional filters shown in FIGS. 7A and 7B can adjust the frequency response in the fx and fy directions, but it is difficult to adjust the frequency response in the oblique direction fxy. Specifically, it is known that when a vertical and horizontal (0 degree / 90 degree direction) one-dimensional filter is used, the frequency response in the fxy direction is the product of the frequency response of fx and the frequency response of fy. The filters in FIGS. 7A and 7B are filters that correct sharpness deterioration, and therefore the frequency response value is 1.0 or more, and the higher the frequency region, the larger the value. As a result, as shown in FIG. 9C, the frequency response value of fxy has a larger response than that of fx and fy, and the magnitude of this frequency response value is particularly remarkable in the high frequency region. End up. As a result, problems such as noise amplification, ringing, and color misregistration are likely to occur.

  Next, consider a case where correction (recovery) is performed with two one-dimensional filter coefficients used in the first embodiment with reference to FIG. In the horizontal and vertical one-dimensional filter (product-sum operation) processing shown in FIGS. 7A and 7C, the frequency response in the fx / fy directions can be adjusted. It is difficult to adjust the frequency response.

  However, as described above, the filter coefficient in the fy direction is ½ in terms of resolution. For this reason, the frequency response in the high frequency component has a shape in which aliasing occurs and the high frequency response value decreases. As a result, as shown in FIG. 10C, the frequency response in the fxy direction shown as the product of the frequency response of fx and the frequency response of fy has a high frequency response value as compared with the conventional case (FIG. 9). Will be suppressed.

  As described above, in the filter processing according to the first embodiment, it is possible to suppress overcorrection in the high-frequency component in the fxy direction. As a result, noise amplification, ringing, color misregistration, and other adverse effects are unlikely to occur. The response value of the high-frequency component in the fy direction is less than 1.0, but as described above, since it is a component region that is insensitive to human vision, it is unlikely to cause a problem.

Increase rate of the total number of transfer pixels For the same reason as the conventional process described with reference to FIG. 4E, there are areas in the periphery of the digital image data 401 and 402 that cannot be subjected to the spatial filter process. To do. In order to perform image processing without any gap between the band regions (band regions 301 and 302), overlapping pixel regions 460 as shown in FIG. 6E are essential.

  However, in the example (fh = 3, Bdh = 8) in the first embodiment (FIG. 6), it can be seen that an 8-line band memory is sufficient to process 6 lines. That is, in the first embodiment, when the increase rate r of the total number of transfer pixels is calculated using Equation (1), r = 1.33, which is 1.33 times the transfer of the actual effective processing area. You can see that it is over That is, the conventional variable separation filter requires an 8-line band memory to process 4 lines, and the increase rate r of the total transfer pixel number is r = 2. It turns out that it can suppress.

  From equation (1), it can be seen that when Bdh is constant, the increase rate r decreases as the filter height fh decreases and approaches “1”. Therefore, it can be seen that the total number of transfer pixels is reduced in the first embodiment as compared with the conventional variable separation filter.

  As described above, according to the first embodiment, the image data thinned out every other row is processed by the variable separation filter. In particular, among the one-dimensional filters in two directions, a filter coefficient with reduced resolution is used for the thinned direction. With this configuration, it is possible to achieve a reduction in overcorrection in the oblique direction and a reduction in the total number of transfer pixels as compared with the conventional variable separation filter.

(Modification)
In the above description, an example has been shown in which even-numbered row images 401 and odd-numbered row images 402 are sequentially developed in a band memory and individually subjected to filter (product-sum operation) processing. Specifically, as shown in FIGS. 6A to 6D, the main scanning one-dimensional filter processing and the sub-scanning one-dimensional filter processing are performed twice for each of the even-numbered row image and the odd-numbered row image (four times in total). The example of performing the filter (product-sum operation) processing of) is shown.

  However, as shown in FIG. 5D, the even-numbered row image 401 and the odd-numbered row image 402 may be developed in the band memory as data of one band region. Specifically, a band memory may be developed in which an even-numbered row image 401 of processing is arranged on the left side and an odd-numbered row image 402 is arranged on the right side. At this time, the invalid image region 403 may be inserted in the central region so that there is no mutual influence between the processing for the even-numbered row image 401 and the processing for the odd-numbered row image 402. If such an invalid image area 403 is used, it is not necessary to perform special processing at the left end of the even-numbered row image 401 and the right end of the odd-numbered row image 402.

  In this way, the expansion to the band memory can be done only once. In addition, since even-numbered image and odd-numbered row image are developed as a single image, it is only necessary to perform a total of two times of filter (product-sum operation) processing: main scanning one-dimensional filter processing and sub-scanning one-dimensional filter processing. .

In the first embodiment, an example is shown in which the band memory height Bdh = 8, the conventional variable separation filter height fh = 5, and the filter height fh = 3 of the present invention. However, it is not limited to these values. For example, Bdh = 16, the height fh = 9 of the conventional variable separation filter, and the height fh = 5 of the filter of the present invention may be used. Again,
Conventional: r = 16 ÷ (16− (9-1)) = 2
The present invention: r = 16 ÷ (16− (5-1)) = 1.33
Thus, the total number of transfer pixels can be reduced by 66%.

  In order to efficiently reduce the number of transfer pixels, an effect can be easily obtained in consideration of the length direction of the band memory. Specifically, it is preferable to determine the thinning direction in consideration that the smaller Bdh can suppress the increase rate r of the total transfer pixel number in Expression (1).

  In the above description, an example of reducing (thinning out) one of the vertical and horizontal resolutions has been described. However, processing for reducing (thinning out) both vertical and horizontal resolutions may be performed. In this way, a further effect can be obtained in suppressing the increase rate r of the total transfer pixel number.

  FIG. 12 is a diagram for explaining a mode in which the resolution is reduced to 1/2 in both of the two one-dimensional filters. That is, in this case, four images 401 to 404 are obtained by the thinning process. Each of the four images 401 to 404 is processed using two vertical and horizontal filter coefficients whose resolution is reduced to ½.

  That is, K divided images obtained by extracting the pixel groups for each L column with different phases are extracted from the K divided images by extracting the pixel groups for each L column with different phases. It may be divided into images. In this case, two one-dimensional filters are sequentially applied to each of the K × L secondary divided images. Then, the K × L post-processed secondary divided images to which the filter process has been applied are synthesized to generate a post-filter process image for the first band region 301.

(Second Embodiment)
In the second embodiment, another form of thinning out in the sub-scanning direction will be described. That is, in the above-described first embodiment, the one-dimensional filter process is performed on an image every other row in the sub-scanning direction (that is, an image having ½ resolution in the sub-scanning direction). However, it is not limited to 1/2 resolution.

  FIG. 11 is a diagram for explaining a mode in which the resolution is reduced to ¼ in one of the two one-dimensional filters. That is, the form which performs a one-dimensional filter with respect to the pixel column for every 4 rows is shown as an example. Specifically, with respect to the input image 400, an image 401 based on the pixel column of the (n + 1) th row (n is an integer of 0 or more), an image 402 based on the pixel column of the (n + 2) th row, and the (n + 3) th row. An image 403 based on the pixel column and an image 404 based on the pixel column on the (n + 4) th row are generated. Then, one-dimensional filter processing in two directions (0 degree and 90 degree directions) is performed on each of the images 401 to 404. At this time, a filter whose resolution in the resolution reduction direction (height direction) is 1/4 is also used as the filter coefficient in the one-dimensional filter processing.

  A case where the height in the band memory is 16 and the filter height fh = 9 of the conventional variable separation filter will be described. At this time, the filter height according to the present invention in which the spatial range exerted by the filter coefficient of the conventional variable separation filter is the same is fh = 3. That is, when the resolution is 1/4 (4 line intervals), if the filter height is fh = 3, the control of the low frequency component is substantially equivalent to the filter height fh = 9 of the conventional variable separation filter. It becomes.

  Since the resolution in the height direction is 1/4, the control performance of the high-frequency component in the height direction is worse than that of the conventional or variable separation filter of the first embodiment. However, due to the human visual characteristics described in the first embodiment, it can be expected that an image that is felt to be almost the same as the conventional filter height fh = 9 is obtained for humans.

At this time, the increase rate r of the total transfer pixel number is as follows.
Conventional: r = 16 ÷ (16− (9-1)) = 2
The present invention: r = 16 ÷ (16− (3-1)) = 1.14
That is, the increase rate can be suppressed by 80% or more. Further, overcorrection in the high frequency region can be further suppressed as compared with the first embodiment.

  As described above, in the second embodiment, it is possible to further suppress the increase rate r of the total number of transfer pixels compared to the first embodiment.

(Third embodiment)
In the third embodiment, a mode in which two one-dimensional filters that are oblique to the main scanning direction are used will be described. That is, in the above-described embodiment, the one-dimensional filter in two directions along the pixel array is used in the input image data. However, a one-dimensional filter in other directions can also be used. In the following description, two orthogonal one-dimensional filters having inclinations of 45 degrees and 135 degrees with respect to the arrangement direction (main scanning direction) of pixel groups in the row direction (or column direction) in the input image data are used. An example will be described.

  FIG. 13 is a diagram illustrating a configuration of a one-dimensional filter in the third embodiment. That is, two one-dimensional filters set in a 45 degree direction with respect to the pixel arrangement are used. Specifically, filter coefficients (FIGS. 13 (c) and (d)) whose resolution is reduced with respect to the conventional filter coefficients (FIGS. 13 (a) and (b)) are used. By performing filter processing using this filter coefficient for each of the four band images that have undergone resolution reduction (decimation) as shown in FIGS. 12B to 12E, the increase rate r of the total number of transfer pixels can be obtained. It becomes possible to suppress. Specifically, the increase rate r of the total transfer pixel number is as follows.

Conventional: r = 8 ÷ (8− (5-1)) = 2
The present invention: r = 8 ÷ (8− (3-1)) = 1.33
That is, the increase rate can be suppressed by 60% or more.

  By the way, by setting the filter in the diagonal direction, it is possible to adjust the frequency response in the diagonal direction (fxy and fyx) direction, while it is difficult to adjust the frequency response in the vertical and horizontal directions (fx and fy). However, the response value in the high-frequency region in the vertical and horizontal directions (fx and fy) is suppressed by using the filter coefficient in which the resolution in the oblique direction is reduced. Therefore, noise amplification, ringing, color misregistration, and the like can be suppressed.

  In the above description, the directions of the two one-dimensional filters are set to be 45 degrees and 135 degrees with respect to the main scanning direction, respectively. In other words, the example in which the two are orthogonal to each other has been described. However, the two one-dimensional filters used need not be orthogonal. For example, filters in two directions (FIGS. 7C and 13C) having inclinations of 90 degrees and 135 degrees, respectively, with respect to the main scanning direction may be used.

  As described above, in the third embodiment, two one-dimensional filters that are oblique to the arrangement of pixels in the image data to be processed are used. As a result, the frequency characteristics in the oblique direction can be directly controlled, and the increase rate r of the total transfer pixel number can be suppressed as compared with the conventional case.

(Other examples)
The present invention supplies a program that realizes one or more functions of the above-described embodiments to a system or apparatus via a network or a storage medium, and one or more processors in a computer of the system or apparatus read and execute the program This process can be realized. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

  110 CPU circuit unit; 120 image reading unit; 130 image processing unit; 140 printer unit; 301, 302 band region; 410, 412 filter coefficient

Claims (10)

An image processing apparatus that divides an input image into a plurality of band area images and sequentially stores each band area image in a band memory to perform image processing,
A first filter coefficient having a reference region of M pixels in the first direction and a single pixel in a second direction different from the first direction in the input image, one pixel in the first direction, and the second Storage means for storing a second filter coefficient having a reference area of N pixels (where N <M) in the direction of
Dividing means for dividing the band area image stored in the band memory into K divided images by extracting pixel groups for each K rows at different phases;
Spatial filter means for sequentially applying a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K divided images;
Generating means for synthesizing K post-processed divided images to which the filter processing by the spatial filter means is applied, and generating a post-filtering image for the band region image;
An image processing apparatus comprising: The dividing unit further divides the K divided images into K × L secondary divided images by extracting pixel groups for each of the L columns at different phases,
The spatial filter means sequentially applies a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K × L secondary divided images. ,
The said generating means synthesize | combines the KxL post-process secondary divided image to which the filter process by the said spatial filter means was applied, The post-filter process image with respect to the said band area image is produced | generated. The image processing apparatus according to 1. The first direction is an arrangement direction of pixel groups in a row direction in the input image, and the second direction is an arrangement direction of pixel groups in a column direction in the input image. The image processing apparatus according to 1 or 2. The first direction is an arrangement direction of pixel groups in a column direction in the input image, and the second direction is an arrangement direction of pixel groups in a row direction in the input image. The image processing apparatus according to 1 or 2. An image processing apparatus that divides an input image into a plurality of band area images and sequentially stores each band area image in a band memory to perform image processing,
A first filter coefficient having a reference region of M pixels in the first direction and a single pixel in a second direction different from the first direction in the input image, one pixel in the first direction, and the second Storage means for storing a second filter coefficient having a reference region of M pixels in the direction of
Dividing means for dividing the band area image stored in the band memory into K divided images by extracting pixel groups for each K rows at different phases;
Spatial filter means for sequentially applying a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K divided images;
Generating means for synthesizing K post-processed divided images to which the filter processing by the spatial filter means is applied, and generating a post-filtering image for the band region image;
Have
At least one of the first direction and the second direction is a direction having an inclination with respect to an arrangement direction of pixel groups in a row direction in the input image. The first direction is a direction having an inclination of 45 degrees with respect to the arrangement direction of the pixel groups in the row direction in the input image, and the second direction is an arrangement of the pixel groups in the row direction in the input image. The image processing apparatus according to claim 5, wherein the image processing apparatus is in a direction having an inclination of 135 degrees. The first direction is a direction having an inclination of 90 degrees with respect to the arrangement direction of the pixel groups in the row direction in the input image, and the second direction is an arrangement of the pixel groups in the row direction in the input image. The image processing apparatus according to claim 5, wherein the image processing apparatus is in a direction having an inclination of 135 degrees. A control method for an image processing apparatus that divides an input image into a plurality of band area images and sequentially stores each band area image in a band memory to perform image processing,
A first filter coefficient having a reference region of M pixels in the first direction and a single pixel in a second direction different from the first direction in the input image, one pixel in the first direction, and the second A storage step of storing a second filter coefficient having a reference area of N pixels (where N <M) in the direction of
A dividing step of dividing the band region image stored in the band memory into K divided images by extracting pixel groups for each K rows at different phases;
A spatial filter step of sequentially applying a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K divided images;
Generating a filtered image for the band region image by combining K post-processed divided images to which the filtering process by the spatial filtering step is applied;
A control method for an image processing apparatus. A control method for an image processing apparatus that divides an input image into a plurality of band area images and sequentially stores each band area image in a band memory to perform image processing,
A first filter coefficient having a reference region of M pixels in the first direction and a single pixel in a second direction different from the first direction in the input image, one pixel in the first direction, and the second A storage step of storing a second filter coefficient having a reference area of M pixels in the storage direction in a storage unit;
A dividing step of dividing the band region image stored in the band memory into K divided images by extracting pixel groups for each K rows at different phases;
A spatial filter step of sequentially applying a one-dimensional filter using the first filter coefficient and a one-dimensional filter using the second filter coefficient to each of the K divided images;
Generating a filtered image for the band region image by combining K post-processed divided images to which the filtering process by the spatial filtering step is applied;
Have
At least one of the first direction and the second direction is a direction having an inclination with respect to an arrangement direction of pixel groups in a row direction in the input image.   The program for functioning a computer as each means of the image processing apparatus of any one of Claims 1 thru | or 7.