JP2008192073A - Image processing device and control method therefor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To rotate, zoom, and deform an image at low cost by matrix operation based on a forward transform, in which coordinates on the output side are found from coordinates of an image on the input side, by means of a low-capacity memory. <P>SOLUTION: The image processing device is provided with a first buffer control part controlling a reading buffer storing partial images, a transformation part sequentially transforming the partial images stored in the reading buffer to the positions of the output coordinate, and a second buffer control part controlling a writing buffer storing the partial images transformed to the positions of the output coordinate by the transformation part. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an image processing technique.

In order to process the image data in the raster scanning order on the output side of the image processing, it is necessary to randomly access the image data on the input side (this conversion method is called an inverse conversion method). As a technique for performing image processing by performing inverse transformation from the output coordinate system to obtain a processing region on the input coordinate system, there is one disclosed in Patent Document 1, for example. In Patent Document 1, reverse conversion and forward conversion are described as selection of conversion methods.
JP 11-154228 A

  However, it is not always possible to perform either inverse transformation or forward transformation by selection. This is because in the case of affine transformation using a general linear transformation formula, the transformation matrix does not necessarily have an inverse matrix. When performing inverse transformation, it is necessary to include processing when there is no inverse matrix, and processing time is changed by switching processing only at that time. There is a problem with the conversion process.

  Further, as shown in FIG. 18, as a case where conversion is performed such that a part of an image is extracted while being rotated, for example, a case where a region 1802 in an input image 1801 is rotated and output is considered. When the buffer size that once enters the output buffer is an area 1803, the size of the area necessary as input data including the width size 1804 and the height size 1805 is indicated as a height size 1806. Thus, the larger the rotation angle, the larger the range (size) of the input image that must be held at one time.

  In Japanese Patent Laid-Open No. 2004-228688, a technique for inputting a necessary image as an input in units of lines and performing a process of accumulating input line image data until data sufficient to read local data in a range as indicated by reference numeral 1807 in FIG. 18 is collected. Is disclosed. Therefore, a large input buffer size is required for the rotation scaling process.

  For example, if the horizontal width of the image is assumed to be A4 size 600 dpi, one line is 5000 pixels or more, so even if a 45 degree rotation is assumed, 5000 / √2, that is, 3535 lines of data must be retained. Don't be. In this case, if one pixel is held in 3 bytes, the memory amount is 5000 × 3 × 3535 = about 50 MB, and a huge memory is required.

  In order to avoid such a problem, it is conceivable to compress and hold the image on the input side. If a compression method with a compression ratio of 10 is used, 50 MB data can be reduced to a relatively realistic size of 5 MB. However, in this case as well, in the inverse conversion method, the input data must be read so as to cross diagonally, so that the reading of the compressed data is reversed. As a result, it takes too much time for reading, and the processing can be performed at high speed. There is a problem that you can not. In particular, in an image compression method such as JPEG, PNG, TIFF, etc., which compresses while moving in the column direction in units of pixels or blocks, the load for reading obliquely is very large and is difficult to use in practice. There is a problem.

  Furthermore, a technique for performing rotation using an address conversion ROM is disclosed in Patent Document 1, and in the case of performing affine conversion, in addition to rotation / magnification, deformation in a form that maintains parallelism is possible. In order to deal with all cases such as rotation and scaling, the address conversion table must be stored in ROM for all combinations of individual coefficients of the 2 × 2 or 3 × 3 conversion matrix. Has a problem that it is difficult to realize.

  In recent standardized SVG (Scalable Vector Graphics) standards and the like, it is possible to freely specify the coefficients of the affine transformation matrix in addition to rotation and scaling. In application to the SVG standard, there is a problem that the ROM table is not necessarily an effective means for executing processing at low cost and at high speed.

  In view of the above problems, the present invention can perform rotation / magnification / deformation processing at high speed and inexpensively by a matrix operation based on a forward conversion method that obtains output-side coordinates from input-side image coordinates using a memory with a small capacity. An object is to provide an image processing technique to be realized.

In order to solve the above-described conventional technology, an image processing apparatus according to the present invention includes first buffer control means for controlling first storage means for storing a partial image, and
Conversion means for sequentially converting the partial images stored in the first storage means into positions of output coordinates;
Second buffer control means for controlling second storage means for storing the partial image converted to the position of the output coordinate by the conversion means;
It is characterized by providing.

  According to the present invention, image processing that realizes rotation / magnification / deformation processing at high speed and at low cost by a matrix operation based on a forward conversion method that obtains output side coordinates from input side image coordinates using a small capacity memory. Technology can be provided.

  Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings. However, the constituent elements described in this embodiment are merely examples, and the technical scope of the present invention is determined by the scope of claims, and is limited by the following individual embodiments. is not.

(First embodiment)
FIG. 1 is a diagram illustrating components included in an image processing apparatus according to an embodiment of the present invention. The image input unit 101 inputs image data and coordinate information of the image data. The image input unit 101 can also acquire image data and coordinate information by simple memory access, hard disk access, or network access.

  The configuration of the image processing apparatus may be a configuration in which a compressed image expansion unit 1501 is provided in addition to the image input unit 101 that acquires image data. According to the configuration shown in FIG. 15, the compressed image decompression unit 1501 can decompress and output the compressed image before the rotation / magnification / deformation processing. In that case, since rotation, scaling, and transformation processing can be performed in units of several pixels at most, a partial image in pixel (pixel) unit or block unit, and its coordinate information, without a special buffer. It is also possible to perform the delivery of. For example, DCT, DFT, DST, LOT, Hadamard transform, Wavelet transform, subband coding, run length coding, LZW coding, Huffman coding, or a combination thereof can be used as the coding method. .

  The read buffer 102 is a buffer that holds the image data input from the image input unit 101. The read buffer 102 can function as a first storage unit that stores partial images.

  The read buffer control unit 103 controls the read buffer 102 (first storage means) that stores local data (partial image) of the input image data. The read buffer control unit 103 can function as a first buffer control unit that controls the read buffer 102 (first storage unit).

  The read buffer control unit 103 deletes unnecessary image data so that image data necessary for processing is stored as local data (partial image) in the read buffer 102 (first storage unit), or creates a new image. Control to hold data is possible. Further, the read buffer control unit (first buffer control unit) 103 can perform rearrangement such as shifting the holding position of the image data in the read buffer 102 (first storage unit) as necessary. It is.

  The local data conversion unit 104 performs coordinate conversion for sequentially converting the local data (partial image) of the input image rearranged by the read buffer control unit (first buffer control unit) 103 to the position of the output coordinate by forward conversion. Can be output.

  The local data output address calculation unit 105 can calculate output coordinate points necessary for coordinate conversion by forward conversion from local data (partial image).

  The forward conversion coordinate calculation unit 106 can obtain the output coordinate position from the local data output address calculation unit 105 by the forward conversion method based on the input coordinate position and the conversion matrix.

  In general image processing, image data is processed in the order of raster scanning, and the image processing results are output in the order of raster scanning. On the other hand, in the rotation process, the order of data input / output differs between the input side and the output side. Image data is input from the input side in the order of raster scanning, and the rotated image data is written in the direction of inclination according to the rotation angle. For this reason, on the output side, access (random access) to the write buffer 107 different from the input order is required. This conversion method is hereinafter referred to as a forward conversion method.

  The write buffer 107 holds the output image data converted by the local data conversion unit 104. The partial image (local data) stored in the read buffer 102 (first storage unit) is converted into the output coordinate position by the local data conversion unit 104 and stored in the write buffer 107. The writing buffer 107 can function as a second storage unit that stores the partial image converted into the position of the output coordinate.

  The write buffer control unit 108 controls a write address for efficiently holding output data in the write buffer 107 (second storage unit). The write buffer control unit 108 can function as a second buffer control unit that controls the second storage unit that stores the partial image converted into the output coordinate position.

  The image output unit 109 appropriately outputs the output image data held in the write buffer 107 (second storage unit) to an external processing module or an external memory.

(Local data conversion by forward conversion)
A local data conversion process by forward conversion will be described with reference to FIG. In FIG. 2, a data flow is assumed in which an input image is sequentially input in a horizontal direction indicated by an arrow 250 in units of pixels (pixels). As shown in FIG. 2A, the number of processing is reduced by processing four pixels at a time, and reading / writing on the read buffer 102 (first storage unit) and the write buffer 107 (second storage unit). The number of times can be reduced.

  Output pixels are output while simplifying the processing by limiting the output pixels to be output based on the four pixels 210 to 213 in the horizontal direction to the pixels included in the area indicated by A in FIG. It can reduce the number of reading and writing on the buffer.

  In principle, it is possible to calculate an output value for a range indicated by B in FIG. 2A, which is a range in which all four points (four pixels) are involved. In this case, the ratio of pixel involvement with respect to the output value differs for each output pixel, making it impossible to proceed with the output value calculation process by a uniform process. Therefore, in the present embodiment, it is assumed that output is performed only for the area indicated by A with respect to four inputs (four pixels). “◯” in the figure shows an output pixel by way of example.

  Next, as shown in FIG. 2B, when one next pixel (one pixel) comes in, the first input pixel is discarded from the storage area, and the newly input one pixel is The same output value is calculated by holding 4 points (4 pixels) including the same. And the output to the output coordinate point contained in the range shown by C in FIG.2 (b) is performed. In this way, for the image input of the first raster, output pixels are sequentially output with very small holding data of 4 points for each input of one pixel.

  Next, as shown in FIG. 2 (c), when the processing of the second raster is performed, an area that can be output by four points is also in the upward direction. Output to the output coordinate point.

  Similarly to the first raster, the second raster is discarded when the next pixel is input as shown in FIG. 2D, and the same processing is performed on four new points (four pixels). The process can proceed sequentially.

  Thereafter, as shown in FIGS. 2E and 2F, it is possible to generate an output pixel by sequentially repeating the same processing as the third raster and the fourth raster.

  The lower end of the image is processed so that the region shown in FIGS. 2 (a) to 2 (d) is just flipped up and down, so that the output processing at the boundary portion of the input image is different from the processing other than the boundary. It is possible to do the same. When performing image composition with a fixed transmittance, it is possible to obtain a composite image by reading an image drawn in advance on the output buffer in units of pixels and writing it back with a coefficient determined by the transmittance. Is possible.

  Next, with reference to FIG. 3, the minimum buffer configuration required to reduce the number of times of reading and writing on the buffer in general processing excluding the upper and lower ends of the image when realizing processing at higher speed will be described. .

  For example, the processing of this embodiment can be realized by hardware or a CPU having a local buffer such as a cache memory. In this case, the number of times of writing to the external memory can be effectively reduced by having a local buffer having a size corresponding to the output coordinate point included in the region shown in FIG. In this case, the size of the buffer varies depending on the scaling factor. This is because when the enlargement process is performed, the number of output coordinate points included in the region in the figure increases, and in the case of reduction, the number decreases.

  In general, since the size of the local buffer is limited in terms of mounting cost, it is necessary to limit the scaling factor. However, according to the present embodiment, the output coordinate points of the area that does not enter the local buffer may be directly output to the external memory and read and processed again in the subsequent processing, and all output coordinates are stored in the local buffer. Point information need not be included. When a scaling factor exceeding the upper limit of the local buffer is specified, the processing load only increases as the scaling factor increases, and the speed and flexibility are relatively easy to control. It is possible to achieve both.

  In the configuration of the local buffer, it is necessary to have the width of the local buffer by the number of pixels in the horizontal direction in which pixels are continuously input. A method for effectively reducing this will be described with reference to FIG. Reference numeral 401 in FIG. 4A schematically shows the input order of the input data shown in FIG. When data is processed in units of rasters in this way, the input order of input data is as indicated by reference numeral 401. In this case, the present invention can be applied to many image processes, but on the other hand, the capacity of the buffer to be prepared in the present embodiment becomes large as indicated by reference numeral 402 in FIG.

  On the other hand, when processing an image in band units, the pixel input direction is set to the vertical direction, and the local buffer area required for the rotation process can be rotated by 90 degrees. As a result, the size of the local buffer can be kept small.

  Next, the processing of the local data conversion unit 104 when performing image rotation will be described with reference to FIG. In the rotation, for example, output pixels are arranged obliquely in the region F in FIG. In this case, it is necessary to perform the integration process after calculating each of the points indicated by reference numeral 501, but the process is the same as the process described so far. For example, even with 16-point interpolation processing, it is possible to retain local data of an input image that is very small of 4 pixels (pixels) and to perform efficient rotational scaling.

  However, this alone leaves elements that increase the circuit scale in hardware, such as the need to perform multiple multiplication processes for affine transformation. Therefore, a configuration in which an output coordinate range and a set of output coordinate points to be actually output included in the output coordinate range will be described below with a simple calculation.

  It is known that a transformation by matrix operation called affine transformation has a property that “parallel straight lines remain parallel after transformation”. This indicates that, when a region surrounded by straight lines is deformed, a figure can be converted by converting inclination information by the number of non-parallel straight lines surrounding the region.

  For this reason, the conversion parameter in the affine transformation is obtained as a unit movement amount after conversion with respect to the unit movement amount of the integer position of the input coordinate system obtained from matrix calculation, that is, as a slope.

  The affine transformation is expressed as follows, where the input coordinates are (x, y) and the output coordinates are (X, Y).

  Here, in the digital image processing, since the input / output coordinate position is discretized at an integer position, for example, the point (x, 0) is called (ax + c, dx + f) by this equation Mapped to position.

  Furthermore, the position x + 1 is mapped to the position (ax + c + a, dx + f + d) as shown in the following equation.

  As is clear from the equation (3), the movement to the adjacent pixel position in the input coordinate system is a movement having an inclination of (a, d) in the output coordinate system. In the coordinate transformation performed by the affine transformation, it can be seen that a certain amount of movement in the input coordinate system becomes a movement amount having a converted inclination in the output coordinate system. From this, it can be seen that in the affine transformation, the converted output position may be easily obtained by giving the transformation coordinates of a certain point on the input coordinates and the movement amount therefrom as the inclination.

  In the case of transformation based on a determinant that does not have an inverse matrix, phenomena such as straight lines that are not parallel to each other on the input image become parallel or aggregated at one point. In such a case, an infinite number of input coordinate positions corresponding to a certain output coordinate exist, and therefore it is impossible to obtain the input coordinate position by inverse transformation by simple calculation. Even in such a case, straight lines having different inclinations before conversion have the same inclination after conversion, and the conversion can be performed by conversion of the inclination.

  FIG. 6 is a diagram for explaining a simple method for obtaining coordinate values when rotation, deformation, and scaling are performed by forward conversion. In the coordinate transformation by affine transformation, the input image coordinates are mapped to the output image coordinate system while maintaining the parallelism of the axes. When the coordinate points of the input coordinate system are indicated by white circles, the coordinate points arranged as squares are generally mapped into a parallelogram arrangement in which the coordinate points are indicated by black circles by performing affine transformation. The X coordinate and Y coordinate of the input coordinate system are mapped to straight lines having inclinations of d / a and e / b, respectively, in the output coordinate system.

  As a result, it is possible to simplify the rotation, scaling, and deformation processing of an image having a rectangular area. FIG. 7 is a diagram illustrating setting parameters when performing rotation / magnification / deformation processing by affine transformation.

  In FIG. 7, pixel points on the output coordinates are indicated by black circles, and pixel points on the input coordinate system are indicated by squares. The position (StartX, StartY) where the processing start point of the input image is mapped on the output coordinates is used as the base point.

  The arrangement of the points of the input image on the output coordinates can be obtained by calculating the slopes (rate_x, rate_y) and (line_rate_x, line_rate_y) of the X and Y axes of the input coordinates on the output coordinate system.

  The coordinate calculation after scaling by affine transformation can be replaced by six parameters (StartX, StartY), (rate_x, rate_y), and (line_rate_x, line_rate_y).

  The six parameters are basic for expressing the affine transformation, but it is difficult to determine the pixel position on the output coordinates only by this. This is because the pixel position of the input image is generally mapped to the middle of the pixel position of the output image, that is, a position including a value after the decimal point. In order to obtain the output image correctly, a calculation for obtaining the output coordinate grid point represented by an integer from the position including the decimal point must be performed.

  How to give this parameter will be described with reference to FIG. Straight lines line1, line2, and line3 are obtained from the six parameters given previously. Then, intersections with the horizontal line at the Y coordinate integer position in the output coordinate system ((CstartX1, CstartY1), (CstartX2, CstartY2) and (CendX1, CendY1), (CendX2, CendY2) in FIG. 8) are calculated.

  This is not suitable for low-priced implementation because it requires arithmetic processing with a relatively large circuit scale including a multiplier in order to be performed by hardware. Must not. It is possible to obtain the range of output coordinates sequentially by giving these intersection coordinates as initial values and then adding inclination information. As a result, the forward conversion coordinate calculation unit for affine transformation can be implemented using only an adder with a low hardware implementation load, and high-speed processing can be performed with an inexpensive configuration.

  As shown in FIG. 8, (StartX, StartY), (EndX, EndY), (EndLineX, EndLineY), and (EndRectX, EndRectY) are obtained as the coordinate positions for the four points that are input first (rectangular range 801). Each of these can be obtained by multiplying a desired affine transformation matrix.

  In the case of a filter having n × m reference points, (x + cn1, y-cm1), (x + cn2, y-cm1), (x +) by coefficients cn1, cn2, cm1, and cm2 obtained from n and m cn1, y + cm2), (x + cn2, y + cm2) can be obtained. Basically, cn1, cm1, cn2, and cm2 are simply n / 2-1, m / 2, n / 2, and m / 2.

  However, since the filter sizes n and m are not necessarily even numbers, cn1, cn2, cm1, and cm2 are not always obtained by such a simple expression. Further, at the upper end, the lower end, the left end, and the right end of the input image, it is necessary to perform control to further change the cn1, cn2, cm1, and cm2 as shown in FIG. However, it is also possible to configure such that the output area is not controlled by supplementing the image necessary for processing by supplying a portion that falls outside the image in consideration of n and m as input pixels. It is.

  With respect to the rectangular range 801 thus obtained, the intersection point with the horizontal line (line 5) having an integer position in the output coordinate system closest in the image processing direction with (StartX, StartY) as the base point is obtained. (CstartX1, CstartY1) is an intersection of line 5 and line 5 extending from the (EndLineX, EndLineY) to (rate_x, rate_Y) described above. In this point, the Y coordinate is an integer position on the output coordinate, but the X coordinate is a position having a value after the decimal point. Similarly, (CstartX2, CstartY2) is an intersection of lines line2 and line5 extending from (StartX, StartY) with a slope of line_rate_y / line_rate_x. (CendX1, CendY1) is the intersection of lines line3 and line5 extending from (EndX, EndY) with a slope of line_rate_y / line_rate_x.

  (StartX, StartY), (EndX, EndY), (EndLineX, EndLineY), (EndRectX, EndRectY) are given as initial values for calculating the output coordinate area. After that, by adding the information of the inclination, it becomes possible to obtain the conversion of the local data in the output coordinates sequentially.

  The flow of the output coordinate calculation process of this embodiment is demonstrated with reference to the flowchart of FIG. This process can be executed under the control of the local data conversion unit 104, the local data output address calculation unit 105, and the forward conversion coordinate calculation unit 106.

  When the output coordinate position calculation process is started, in step S901, the input coordinate CstartX1, CstartX2 having the larger X coordinate value is selected and set as CstartX.

  In step S902, of the input coordinates CendX1 and CendX2, the smaller of the X coordinates is selected and set as CendX.

  In step S903, since CstartY1, CstartY2, CendY1, and CendY2 have the same value for all points, any CY is determined.

  In step S904, CstartX is rounded up and moved to the integer position to obtain CX.

  In step S905, the current CX, CY is output as an output coordinate position to the local data output address calculation unit 105 shown in FIG.

  In step S906, 1 is added to CX in the output coordinate system.

  In step S907, if CX is smaller than the previously determined CendX (S907-Yes), the process returns to step S905. On the other hand, if it is determined in step S907 that CX is equal to or greater than Cendx (S907-No), the process proceeds to step S908.

  In step S908, 1 is added to CY.

  In step S909, if CY exceeds EndRectY, the process ends. On the other hand, if it is determined in step S909 that CY does not exceed EndRectY (S909-Yes), the process proceeds to step S910.

  In step S910, by adding rate_y / rate_x to CstartX1, it is moved to the intersection of the horizontal line CY updated in step S908 and line1 in FIG. 8, and CstartX1 is updated.

  In step S911, by adding line_rate_y / line_rate_x to CstartX2, it is moved to the intersection of the CY horizontal line and line2 in FIG. 8, and CstartX2 is updated.

  In step S912, by adding line_rate_y / line_rate_x to CendX1, it is moved to the intersection of the horizontal line of CY and line3 in FIG. 8, and CendX1 is updated.

  In step S913, by adding rate_y / rate_x to CendX2, the CY horizontal line updated in step S908 is moved to the intersection of line4 in FIG. 8, CendX2 is updated, and the process returns to step S901. Then, the same processing is repeated from step S901 to S913.

  According to the present embodiment, an image that realizes rotation, scaling, and transformation processing at high speed and low cost by a matrix operation based on a forward transformation method that obtains output side coordinates from input side image coordinates using a small amount of memory. Processing technology can be provided.

(Second Embodiment)
FIG. 10 shows the configuration of the image processing apparatus according to this embodiment. Since the components (101 to 109) are the same as those shown in FIG. 1, detailed description thereof is omitted here. In the image processing apparatus according to the present embodiment, the data 1010 exchanged between the image input unit 101 and the reading buffer 102 includes not only pixel information but also coordinate information.

  Input image data to be given to image processing is not necessarily in units of one raster. This is because, in addition to the case of inputting one line in the horizontal direction as shown in FIG. 4A, data may be sequentially input in the vertical direction for partial images divided into bands (FIG. 4B). )). In order to adapt to the input order of such various input data, it is necessary to have a data structure in which coordinate information is also included in the data 1010 of FIG.

  FIG. 11 shows an example of a suitable data input order when partial image data (pixel data) and coordinate information are input simultaneously.

  The data input order is to input the two rasters 1101 sequentially in column units. In this case, the data for two rasters can be calculated at once by the equation (4) as compared with the data input of one raster unit already described in the first embodiment. In the equation (4), ci is a coefficient indicating the pixel influence degree at the time of scaling in the horizontal direction. cj is a vertical coefficient, and Pin is a pixel value at a position (i, j) in the local image range. The addition for generating the output image can be calculated by equation (5).

  The number of integrations on the output memory or the output-side local memory can be reduced to half or half + 1. In particular, when the output buffer is placed in an external memory with a slow response speed, it is expected to increase the processing speed. In addition, since the horizontal coefficient ci for one output point can use the same value for two rasters, the processing speed for calculating the coefficient can be improved.

  FIG. 12 shows a local memory area 1201 to be provided on the output side when data is input in units of two rasters.

  In the case of 2-raster input, it can be seen that the overlapping portion of the buffers is smaller than in the case of 1-raster input (FIG. 3). By having a local memory for covering such an area, the integration process can be performed in the local memory, and access to the external memory with a slow response can be suppressed to one write.

  Even when processing a filter size that exceeds the limit of the local memory capacity, processing can be performed by allocating the protruding part to external memory, and the filter size that can be processed is not limited by the size of the local memory .

  FIG. 13 is a diagram illustrating a state in which three rasters are input in column units 1301.

  In this case, the equations (4) and (5) shown above become the following equations (6) and (7).

  The number of integrations in the output memory or the output-side local memory can be reduced to 1/3 or 1/3 + 1. In particular, when the output buffer is placed in an external memory with a slow response speed, a high-speed process can be expected. . Further, since the same value can be used for the three rasters for the horizontal direction coefficient ci for one output point, the processing speed for calculating the coefficient can be further improved.

  In addition, by selecting coefficients ci, cj, filters, and reference points for interpolation calculation, processing that is performed by switching processing such as primary interpolation and tertiary interpolation without changing the configuration of the local data conversion processing, for example. It may be implemented in hardware or software.

  Furthermore, it is possible to add a determination process for determining the nearest point in the step of obtaining the horizontal and vertical distances of the surrounding local data when calculating the coefficients. By adding this determination step, it is possible to implement a process for performing 0th-order interpolation (nearest neighbor sampling process) in hardware or software without changing the configuration of the local data conversion unit.

(Third embodiment)
As a third embodiment, processing when data having the same number of lines as the filter size is simultaneously given as input will be described. In the portion indicated by reference numeral 1401 in FIG. 14, since all input pixels for filter calculation are held at the same time, there is no need for integration. The area indicated by reference numeral 1402 in FIG. 14 does not need to be stored in the local buffer because the integration process divided into a plurality of processes is unnecessary. Here, a special case in which integration processing is not required is described as an example, and a local buffer can be used as a temporary holding buffer for performing continuous writing (burst write) to an external memory. .

(Fourth embodiment)
FIG. 16 is a diagram for exemplarily explaining input / output restrictions on image data. An image area 1601 is an image area showing the entire input image data. An area 1602 indicates an area to be processed in the image area 1601. The result of processing all the images in the area 1602 is stored in a buffer area 1603 that holds data of the output image. Actually, there are cases where the buffer capacity for holding output image data is limited due to a reason that a sufficient capacity cannot be secured.

  FIG. 17 is a diagram illustrating a configuration of the image processing apparatus according to the present embodiment. An input area holding unit 1710 and an output area holding unit 1711 are provided for the configuration of FIG. Note that the image input unit 101 may be provided instead of the compressed image expansion unit 1501.

  The input area holding unit 1710 holds a coordinate range to be processed in the input image. Based on the coordinate range held by the input area holding unit 1710, the read buffer control unit (first buffer control unit) 103 determines whether the coordinates of the input image are included in the area to be processed. When the coordinates of the image input based on the determination result are included in the coordinate range to be processed, the reading buffer control unit 103 stores the data in the reading buffer 102, and reads from the reading buffer 102 to the local data conversion unit 104. Control the output to.

  However, in scaling processing and filter processing that require interpolation calculations, etc., the input image (input pixel) is selected after considering the filter range and the original range around the input area that is referenced during interpolation. Processing is performed by the read buffer control unit 103.

  The output area holding unit 1711 holds information on the coordinate range of the area to be output. The writing buffer control unit (second buffer control unit) 108 determines the position of the partial image (output image) based on the partial image (output image) converted into the coordinate position and the position information 1730 in the output coordinate system. It is determined whether the information is included in the coordinate range of the area to be output. When the position information of the output image is included in the coordinate range of the area to be output based on the determination result, the write buffer control unit (second buffer control unit) 108 displays the output image and the position information in the write buffer 107. And output to the image output unit 109 is controlled.

(Other embodiments)
Needless to say, the object of the present invention can also be achieved by supplying a system or apparatus with a computer-readable storage medium storing software program codes for realizing the functions of the above-described embodiments. Needless to say, this can also be achieved by the computer (or CPU or MPU) of the system or apparatus reading and executing the program code stored in the storage medium.

  In this case, the program code itself read from the storage medium realizes the functions of the above-described embodiments, and the storage medium storing the program code constitutes the present invention.

  As a storage medium for supplying the program code, for example, a flexible disk, a hard disk, an optical disk, a magneto-optical disk, a CD-ROM, a CD-R, a nonvolatile memory card, a ROM, or the like can be used.

  Further, the functions of the above-described embodiment are realized by executing the program code read by the computer. In addition, an OS (operating system) running on a computer performs part or all of actual processing based on an instruction of a program code, and the above-described embodiment is realized by the processing. Needless to say.

It is a figure which shows the component with which the image processing apparatus which concerns on 1st Embodiment is provided. It is a figure explaining conversion processing of local data by forward conversion. It is a figure explaining a local buffer. It is a figure explaining a local buffer. It is a figure explaining the process of the local data conversion part 104 in the case of rotating an image. It is a figure explaining how to obtain | require the coordinate value at the time of performing rotation, deformation | transformation, and scaling by forward conversion. It is a figure explaining the setting parameter in the case of performing the rotation / magnification / deformation processing by affine transformation. It is a figure explaining the setting parameter in the case of performing the rotation / magnification / deformation processing by affine transformation. It is a figure explaining the flow of an output coordinate calculation process. It is a figure which shows the component with which the image processing apparatus which concerns on 2nd Embodiment is provided. It is a figure which shows the example of the suitable data input order in the case of passing over partial image data and coordinate information simultaneously. It is a figure which shows the area | region of the local memory which should have on the output side when the data input per 2 raster units is performed. It is a figure which shows the state which inputs 3 rasters by the column unit 1301 It is a figure explaining the process at the time of giving simultaneously the data of the same number of lines as a filter size as an input. 2 is a diagram illustrating a configuration of an image processing apparatus including a compressed image expansion unit 1501. FIG. It is a figure explaining the input / output restriction | limiting of image data exemplarily. It is a figure which shows the structure of the image processing apparatus which concerns on 4th Embodiment. It is a figure explaining the conversion which takes out part rotating in an image exemplarily.

Claims (8)

First buffer control means for controlling first storage means for storing partial images;
Conversion means for sequentially converting the partial images stored in the first storage means into positions of output coordinates;
Second buffer control means for controlling second storage means for storing the partial image converted to the position of the output coordinate by the conversion means;
An image processing apparatus comprising: An image input means for inputting the partial image and coordinate information of the partial image;
The said 1st storage means stores the said partial image input by the said image input means, and the coordinate information of the said partial image by control of the said 1st buffer control means, The said 1st storage means is characterized by the above-mentioned. Image processing device. A compressed image expansion means for expanding and outputting the compressed image;
2. The first storage unit stores a partial image of an image expanded by the compressed image expansion unit and coordinate information related to the partial image under the control of the first buffer control unit. An image processing apparatus according to 1. An input area holding means for holding information on the coordinate range of the partial image to be processed among the input partial images is further provided.
The first buffer control means determines whether the coordinates of the input partial image are included in the coordinate range of the partial image to be processed based on the information of the coordinate range,
When the coordinates of the input partial image are included in the coordinate range of the partial image to be processed based on the determination, the first buffer control means The image processing apparatus according to claim 1, wherein the image processing apparatus is controlled to write an image. It further comprises output area holding means for holding information on the coordinate range of the area to be output,
The second buffer control unit determines whether the position information of the output image output by the conversion unit is included in the coordinate range of the region to be output,
The second buffer control unit writes the output image in the second storage unit when the position information of the output image is included in the coordinate range of the region to be output based on the determination. The image processing apparatus according to claim 1, wherein the image processing apparatus is controlled as follows. A first buffer control step in which the first buffer control means controls the first storage means for storing the partial image;
A converting step in which the converting means sequentially converts the partial images stored in the first storing means into positions of output coordinates;
A second buffer control step, wherein the second buffer control unit controls a second storage unit that stores the partial image converted into the position of the output coordinate by the conversion step;
An image processing apparatus control method comprising:   A program for causing a computer to execute the control method of the image processing apparatus according to claim 6.   A computer-readable storage medium storing the program according to claim 7.

Images