JP4610450B2 - Image processing apparatus for processing fixed length compressed image and packing data of attribute information - Google Patents

Description

  The present invention relates to an image processing apparatus, and more particularly, to an image processing apparatus that processes image data at high speed by packing image information and attribute information that have been fixed-length compressed in block units.

  There is a standard variable-length compression format such as JPEG as an image data format. They often have less image degradation after decompression than the fixed length compression method. Therefore, a variable-length compression format such as JPEG is suitable for compressing and storing an image that is not used for distribution or paper output. At this time, the attribute information is stored separately from the image. Whether the attribute information is reversibly compressed and stored or is stored as it is is arbitrary.

  In order to rotate an image, for example, as shown in FIG. 17A, an image of 8 pixels in the main scanning direction × 8 pixels in the sub-scanning direction is packed in advance (one block so that it can be read out by one memory access). In other words, the number of memory accesses can be reduced and high-speed rotation processing can be performed. When one pixel is represented by 1 byte (8 bits), 8 pixels is 8 bytes. Normally, when rotating an image, if the image is rotated by 8 × 8 pixels, memory access is required 8 bytes × 8 times. On the other hand, if 8 × 8 pixels are packed in advance, memory access is only 64 bytes × 1 time. Since the memory can be burst-accessed if it is a continuous address, one 16-word burst access as shown in FIG. 17 (c) in contrast to eight 2-word burst accesses as shown in FIG. 17 (b). In the latter access method, which requires only the transfer preparation and end processing, only one memory access time is required.

  If the number of memory accesses is reduced by burst transfer, in addition to shortening the memory access time, there is an advantage that the memory access by other DMACs can be avoided as a system. This merit increases as the amount of information per pixel decreases. As in the case of MFP, in the case of information amount of 1 bit per pixel (CMYK image), if it is rotated 32 x 32 pixels, memory access is required 1 word (4 bytes) x 32 times, but packing 32 x 32 pixels in advance. If so, 32 words (128 bytes) × one access is sufficient.

  In consideration of rotating an image, a fixed-length code is advantageous as a compression method. The fixed length code is compressed to a fixed bit length every several pixels. Since the code after compression has a bit length determined for each unit (called a block), the head address for each block can be specified. Therefore, in the case of a fixed-length code method in which adjacent blocks have no correlation, only necessary blocks can be extracted. In the case of a fixed-length code that has no correlation between adjacent blocks, it is possible to acquire and rotate a block necessary for rotation while leaving the compressed image on the memory, and recompress it back to the memory. Yes, it is possible to increase the use efficiency of the memory.

  For fixed-length compression of an image, a method of compressing several pixels of one line as one block and, for example, as shown in FIG. 18, a two-dimensional image such as 8 × 8 pixels as one block is regarded as one block. There is a way to compress. In order to rotate an image, the latter is more advantageous because the number of memory accesses is reduced. For example, Patent Document 3 discloses a method capable of performing fixed-length encoding and performing rotation processing in units of 90 °. Even in the variable-length code method, rotation can be performed by storing the relationship between the block and the address as in the method disclosed in Patent Document 1, for example.

  By the way, in the image information, there is information other than the pixel color called attribute information. For example, it is a case where each pixel has a flag indicating whether the pixel is a photograph, a character, or a graphic as a parameter for image processing. It is also called separation information or separation signal in the sense of a signal that separates characters and photographs. Although attribute information is held for each pixel, the amount of information is 1 bit or 2 bits for each pixel. Generally, there are the following two types of data formats for images with attribute information. (1) An image is handled as one file as an image, and attribute information is handled as one file as attribute information. (2) One attribute information per pixel is treated as a set and arranged in dot-sequential order. For example, if there is 8-bit attribute information in an RGB (24-bit) image, 32 bits are collectively treated as one pixel. For example, like the method disclosed in Patent Document 2, there is a method of packing an image and attribute information in a certain unit block. The serially input image and attribute information are directly packed and parallelized in units of blocks, rotated, and then converted back to serial and output. This is intended for serial input and output from the video I / F.

  Since the color depth information and attribute information of a pixel have different characteristics, it is more efficient to use different algorithms for compression. For example, for RGB images and text / photo separation information, standard compression schemes such as JPEG have been proposed for RGB, but for text / photo separation information, the same information is arranged, so for example, run-length compression is suitable. ing. Since JPEG handles images, it is irreversible compression in which data loss occurs at the quantization stage because it is sufficient to distract human eyes, but separation information needs to be reversibly compressed. . As shown in FIG. 19, since the character portion and the photograph portion are continuous, the same pattern is repeatedly generated in the separation information, so that it is easy to compress. Thus, it can be seen that the image and the attached information are more efficient when handled separately in consideration of the compression rate and the possibility of missing information.

  FIG. 20 is a functional block diagram of a conventional example of an image rotator. In the read from the memory, the fixed-length code image is m word and the attribute information is n word. Since the fixed-length compressed image and the attribute information are separately stored in blocks of a × a pixels, one block of data can be read with two accesses for the image and the attribute. The data write also requires two accesses. Four DMA controllers are required for pre-rotation image reading and post-rotation image writing, and for pre-rotation attribute reading and post-rotation attribute writing. The image block in FIG. 20 is an RGB image of 8 × 8 pixels, for example.

  A conventional method for rotating an uncompressed image by 90 ° will be described with reference to FIG. When an uncompressed image is rotated by 90 ° or 270 °, it is usually divided into several blocks in consideration of access to the memory. For example, as shown in FIG. 21, the blocks are numbered with 8 pixels × 8 lines as one block. There are two types of rotation order. The first is a method of rotating in the order of A1 block → B1 block → C1 block..., A2 block → B2 block → C2 block. This is tentatively called “rotary lead”. This is a method of reading in accordance with the order of arrangement after rotation and writing in a normal order. The second is a method of rotating in the order of D1 block → D2 block → D3 block..., C1 block → C2 block → C3 block. This is temporarily called a “rotating light”. This is a method of reading in a normal order and writing in accordance with the order of arrangement after rotation.

  Since each block consists of 8 lines, at least 8 memory reads are required for each block. In the case of rotation read, it is possible to directly rasterize using a block-raster conversion buffer without writing back to the memory after rotation. However, this is not possible with rotating lights. Note that, in any case of rotation, the attribute information needs to be rotated separately from the image. In this way, the image can be rotated in units of blocks. A compressed image can also be rotated in a similar manner. Below are some examples of prior art related to this.

  The “image processing apparatus” disclosed in Patent Document 1 requires only a small amount of buffer memory to be used when outputting a composite image, has a short processing time for generating a composite image, and has little image quality degradation. It is a processing device. The input image data is compressed for each block by the image compression / decompression unit. The compressed code data is stored for each block in the code buffer. The compressed code data stored in the code buffer is read by the code buffer control unit in the order of blocks corresponding to the designated editing mode. The image compression / decompression unit decompresses the compressed code data read by the code buffer control unit.

  The “composite processing method of image information” disclosed in Patent Document 2 makes effective use of a memory data bus width larger than the number of image data bits, speeds up memory reading and writing, and allows free selection and setting of an image rotation angle and the like. In this method, the image data and the paired feature signals are rotated while maintaining the relationship, and the memory address control of the rotation processing is simplified. Serial image data is input, a feature signal indicating the feature of the image data is input, and image information composed of the image data and the feature signal is converted into block parallel data of main scanning number × sub-scanning number. Parallel data is stored in one address. The block size is changed according to the data bus width of the storage means. The parallel data read from the storage means is returned to the serial image data and the feature signal.

The “image compression processing device” disclosed in Patent Document 3 is capable of creating an image with good image quality even at an edge portion having a high frequency, and has a code amount fixed for each line, so a table for managing the start address for each line Is an image compression processing apparatus that is low in cost and high in reliability. A plurality of fixed length encoding means converts image data into different code lengths in units of blocks of a plurality of pixels. The total code length generated in a predetermined number of blocks is calculated by the code length total calculation means. The use ratio of the plurality of fixed length encoding means is controlled so that the output value of the code length total calculating means is kept below a predetermined value.
Japanese Patent Laid-Open No. 10-075345 JP 2002-125118 A JP2003-219188

  However, in the conventional compressed image rotation method, since the image and the attribute information are separately packed, there is a problem that the number of memory accesses increases and the memory read / write time for image processing such as rotation becomes long. .

  An object of the present invention is to reduce the number of memory access bursts when image information is input / output to / from a memory by burst transfer in order to solve the above-described conventional problems and perform rotation processing in an image processing apparatus, It is to reduce the memory access time.

The image processing apparatus of the present invention includes a storage unit for storing image data, means for fixed length compressed into blocks in the unit square region the image data, means for blocking the attribute information of the image data in the same block A two-dimensional block data format conversion unit comprising: packing the image data and attribute information compressed in fixed-length blocks in units of the square area and storing them in the storage unit; A unit that blocks data in units of pixels in the line direction and compresses the fixed length, a unit that blocks attribute information of the image data in units of blocks in the same line direction, and a unit that blocks in units of pixels in the line direction Means for packing the image data and attribute information compressed in length into blocks and storing them in the storage unit; Each square area in a block group constituted by a plurality of square area unit blocks in the main scanning direction and the sub-scanning direction stored in the storage unit by the obtained one-dimensional block data format conversion unit and the two-dimensional block data format conversion unit An image rotation unit that rotates the block group by an integral multiple of 90 ° with the unit block orientation fixed, and the packing data stored in the storage unit by the one-dimensional block data format conversion unit is decoded in block units in the line direction. And an expansion unit that converts the data into raster data .

Since the square area unit block image data is packed with fixed-length compressed data and attribute information and stored in the storage unit, and the packed data is read out and rotated, the storage unit is used when rotating the image. A two-dimensional block can be read with a single access to, the number of accesses to the storage unit is reduced, and the access time to the storage unit is shortened.
Also, the image data and attribute information, which are fixed-length compressed in units of multiple pixels in the line direction, are packed and stored in the storage unit, converted to raster data, and image processing is performed. Compressed and packed data can be easily retrieved with a raster and processed on a line-by-line basis.

  Hereinafter, the best mode for carrying out the present invention will be described in detail with reference to FIGS.

  The first embodiment of the present invention is an image processing apparatus that packs and stores fixed length compressed data and attribute information of an image of a square block, reads the packed data, and performs rotation processing.

  The basic configuration of the image processing apparatus according to the first embodiment of the present invention is the same as that of a conventional image processing apparatus. The image processing apparatus according to the first embodiment of the present invention is different from the conventional one in that a fixed-length compressed image and attribute information are packed.

  FIG. 1 is a diagram of a data format used in the image processing apparatus according to the first embodiment of the present invention. This shows a state where one block of fixed-length compressed image and attribute information are packed. FIG. 2 is a functional block diagram of a rotator used in the image processing apparatus. FIG. 3 is a diagram illustrating a method of obtaining raster data from packed data and performing image processing on the raster data. FIG. 4 is a diagram showing a method for printing raster data from packed data. FIG. 5 is a functional explanatory diagram of the block-raster converter. FIG. 6 is a functional block diagram of an image processing module that compresses an uncompressed image or the like with a two-dimensional block and packs it together with attribute information. FIG. 7 is a functional block diagram of a module that performs non-compressed image or the like by fixed length compression in a two-dimensional block, packs it with attribute information, converts it into raster data, and performs image processing. FIG. 8 is a diagram illustrating a method of rotating image data that has been fixed-length compressed in a two-dimensional block and packed together with attribute information.

  2 to 8, the image rotator 1 is means for rotating the entire block by rotating image data in units of blocks in units of pixels. The separating unit 2 is a unit that separates the fixed-length compressed image from the attribute information. The decompressor 3 is means for returning a fixed-length compressed image to an uncompressed image. The image rotation means 4 is means for rotating the image data in units of pixels. The rotation here is a rotation in units of 90 °. The attribute information rotating means 5 is means for rotating attribute information in units of pixels. The fixed length compression means 6 is means for compressing a non-compressed image at a fixed length. The packing unit 7 is a unit that combines the fixed-length compressed image and the attribute information into one data. The block-raster converter 8 is means for generating raster data from image data in units of blocks. The image processing unit 9 is a means for generating new raster data by subjecting raster data to image processing. The plotter 10 is means for printing in raster data units. The CPU 11 is an arithmetic processing unit that executes an image processing program. The memory 12 is a means for storing an image processing program and image data. The data format converter 13 is means for converting an uncompressed image or the like into packing data of a fixed-length compressed image in block units and attribute information. The decoding rotator 14 is means for decoding packing data of fixed-length compressed images and attribute information in units of blocks and rotating the images.

  The function and operation of the image processing apparatus according to the first embodiment of the present invention configured as described above will be described. First, with reference to FIG. 1, a method of packing image information that has been subjected to fixed-length compression and attribute information and storing them in a memory will be described. The image information of one rectangular block is compressed at a fixed length. The rectangular block is, for example, an 8 × 8 pixel block on the screen. The rectangular block may be a rectangular block or a square block. The attribute information of the pixels of this block is collected and stored at the address immediately after the fixed-length compressed image information. The attribute information may be packed in a two-dimensional block. By repeating this, one block of image information and attribute information is continuously stored for each fixed address of the memory.

  Next, the operation of the image rotator used in the image processing apparatus will be described with reference to FIG. The data size of one block of fixed-length compressed image is m words. The data size of the attribute information corresponding to it is n words. m and n are constant values determined by image characteristics and a fixed length compression method. Therefore, the read from the memory is (m + n) words per access. Since the fixed-length compressed image and attribute information are divided into blocks of a × a pixels, one block of data can be read with one access. a is, for example, 8 or 16. Data write is also possible with one access. Two DMA controllers are required for reading data before rotation and for writing data after rotation. The image block in FIG. 2 is, for example, an 8 × 8 pixel RGB image.

  In the image rotator 1, first, the separation unit 2 separates the fixed length compressed image and the attribute information. Only the attribute information is rotated by the attribute information rotating means 5. On the other hand, the fixed-length compressed image is decoded by the decompressor 3 to be an uncompressed image. This uncompressed image is rotated by the image rotation means 4. The rotated uncompressed image is fixed-length compressed by the fixed-length compression means 6. The fixed length compressed image and the attribute information are packed by the packing unit 7. The format and size of the packed data is equal to the format and size of the packing data before rotation. In this way, the image rotator 1 rotates the fixed-length compressed image of one block and the attribute information.

  Next, a method of obtaining raster data from image data obtained by fixing a square block image at fixed length and attribute information and performing image processing on the raster data will be described with reference to FIGS. 3, 4, and 5. As shown in FIG. 3, when image processing is performed on raster data in fixed-length compressed image data that has been blocked in units of a × a pixels, a block-raster converter is required after decompression. First, the fixed-length compressed image data blocked in units of a × a pixels is decoded by the decompressor 3 to be an uncompressed image. The block-raster converter 8 converts an uncompressed image that has been blocked in units of a × a pixels into raster data using a buffer for a lines. The block-raster converter 8 will be described later. The raster data is processed by the image processing unit 9 to obtain raster data of the processed image.

  Also, as shown in FIG. 4, when printing fixed-length compressed image data blocked in units of a × a pixels, a block-raster converter is required after decompression. First, the fixed-length compressed image data blocked in units of a × a pixels is decoded by the decompressor 3 to be an uncompressed image. The block-raster converter 8 converts an uncompressed image that has been blocked in units of a × a pixels into raster data using a buffer for a lines. This raster data is printed by the plotter 10.

  In the block-raster converter 8, as shown in FIG. 5, in order to change a block to a raster, it is necessary to have a line buffer corresponding to one page width for at least a line. Since the image data is compressed and packed in units of blocks, the one-block image after decompression is the shaded block in FIG. Data decompressed in units of blocks is written to the buffer in units of square blocks. After writing the data for a line, read it line by line. In order to perform block-raster conversion at high speed, it is necessary to provide a line buffer of a × 2 lines and perform toggle processing. However, the larger the width of one page, the larger the hardware size and the higher the cost. Therefore, when high-speed processing is not required, the cost is reduced without using a replacement buffer.

  Next, a hardware module for compressing fixed-length image information of a two-dimensional block (square block) and packing it with attribute information will be described with reference to FIG. The variable-length code image shown as an example of the input image data is an image that is variable-length compressed every 8 × 8 pixels (16 × 16 pixels when sub-sampled), such as JPEG. The input image may be an uncompressed image or a fixed-length compressed image in addition to the variable length code. The input image is input by some means, converted back to an uncompressed image by the data format converter 13, and stored in the memory 12 together with the attribute information. The input image is compressed by the data format converter 13 for each 8 × 8 pixel block, packed with attribute information, and written to the memory 12 in the format shown in FIG.

  Converting a variable-length code compressed in block units (such as JPEG) into a fixed-length code compressed in block units, adding attribute information in block units, packing them, and storing them in memory gives the following two advantages Is obtained. That is, (1) when rotating an image, it is not necessary to develop it into an uncompressed image, and the amount of memory can be reduced. Furthermore, (2) the number of memory accesses is reduced, the memory access is accelerated, and the memory access of other devices is not hindered. By the way, the variable length code (JPEG) can be rotated while being decoded, but cannot be written again as the variable length code (JPEG) when writing to the memory. However, it is possible to write back an uncompressed image or a fixed-length compressed image. When an uncompressed image is stored in the memory from the beginning, or when it is stored in the memory with a fixed-length code having no correlation between adjacent blocks, it can be rotated as it is. If the format conversion is performed on the fixed-length code and the packing data of the attribute information, there is an advantage that the rotation process can be easily performed later.

  Next, referring to FIG. 7, the image information of the two-dimensional block (square block) is fixed-length compressed, packed with attribute information, the packed data is rotated and converted into raster data, and the raster data is converted into image data. The hardware module to be processed will be described. 6 is the same as the module of FIG. 6 until the input image is fixed-length compressed for each 8 × 8 pixel block by the data format converter 13, packed together with the attribute information, and written to the memory 12. The packed data is decoded by the decoding rotator 14 into an uncompressed image, and further rotated to obtain a rotated uncompressed image. A specific rotation method will be described later. The rotated uncompressed image is converted into raster data by the block-raster converter 8. The raster data is processed by the image processing unit 9 to obtain raster data of the processed image.

  If the fixed-length code and the packing data of the attribute information are stored in the memory 12, the head address of each block and the data length of the block are known. Therefore, data can be read from the memory 12 for each block so as to rotate in the direction of 90 °, 180 °, or 270 °. If the block is rotated by the decoding rotator 14 and converted to a raster, it can be input directly to the image processing unit 9 for processing in line units without writing back to the memory 12. Advantages in this case are (1) 90 ° or 270 ° image rotation is possible without placing an uncompressed image in the memory, and the cost is low. (2) Since the fixed-length code and the attribute information are packed, the number of memory accesses is reduced and the performance is improved. In particular, since data can be converted into a raster without being written back to the memory 12 after the decoding rotation and input to the image processing unit 9, there is a merit that the memory capacity can be reduced and the performance is remarkably improved.

  Next, with reference to FIG. 8, a method of rotating the blocked image data by 90 ° will be described. Image data and attribute information that have been fixed-length-compressed for each square block of 8 pixels × 8 lines are packed into one data and stored continuously in the memory. As with conventional uncompressed images, “rotational read” (read with a rotated address) and “rotate write” (write with a rotated address) are possible. Since the data of each block is already packed, an 8 × 8 pixel block can be read by one memory read access. The data length of each block varies depending on the algorithm of fixed length compression and the size of attribute information. Due to the fixed length compression, the start address of each block is determined, and the start address of the next block is determined depending on the data length of each block. In consideration of the data length per block, “rotational read” (a method of reading in accordance with the order of arrangement after rotation) or “rotation write” (a method of writing in accordance with the order of arrangement after rotation) is performed. For this purpose, memory read and memory write address generation is performed. In the case of FIG. 7, it is necessary to perform rotation read, and rotation write cannot be performed.

  In the case of “rotational read”, a memory read address and a memory write address are generated as follows. First, in order to read the A1 block, the head address of the A1 block is obtained. (Total start address) + (Block size) × (Number of blocks in one line) × (Number of blocks in one column−1) Read and rotate the A1 block, and write the rotated A1 block to the start address of the rotated memory area. The order of writing is the same as the normal raster order. Next, the B1 block is read. This address is obtained by (start address of A1 block) − (block size) × (number of blocks in one line). The B1 block is rotated, and the rotated B1 block is written to the next address of the A1 block in the memory area after the rotation. Repeat this. When the rotation of one row is finished, read the A2 block. The start address of the A2 block is obtained by (start address of the A1 block) + (block size). The rotated A2 block can be simply written to the previous address.

  In the case of “rotary write”, a memory read address and a memory write address are generated as follows. The reading order is the same as the normal raster order. That is, the D1 block is read first, and thereafter the blocks are read one after another in raster order. Rotate the D1 block and write the rotated D1 block to the corresponding address in the rotated memory area. The write address of the D1 block is obtained by (start address of write area) + (block size) × (number of blocks in one column−1). The D2 block after rotation is obtained by (start address of D1 block) + (block size) × (number of blocks in one column). This is repeated, and when the rotation for one line is completed, the C1 block is read. The write address of the C1 block is obtained by (the start address of the D1 block) − (block size).

  As described above, in the first embodiment of the present invention, the image processing apparatus packs and stores the fixed length compressed data and attribute information of the square block image, reads the packed data, and performs the rotation process. Therefore, when rotating the image, a two-dimensional block can be read with one memory access, the number of memory accesses is reduced, and the memory access time is shortened.

  The second embodiment of the present invention is an image processing apparatus that packs image data and attribute information that have been fixed-length compressed in units of 8 pixel blocks in the line direction, stores them in a memory, converts them into raster data, and performs image processing. is there.

  The basic configuration of the image processing apparatus according to the second embodiment of the present invention is the same as that according to the first embodiment. The image processing apparatus according to the second embodiment of the present invention differs from the first embodiment in that the input image is fixed-length compressed in units of a plurality of pixels in the line direction and packed together with attribute information thereof.

  FIG. 9 is a functional block diagram of a rotator used in the image processing apparatus according to the second embodiment of the present invention. FIG. 10 is a diagram illustrating a case where raster processing is obtained from packed data and image processing is performed by the image processing apparatus. FIG. 11 is a diagram illustrating a case where raster data is obtained from packed data and printed by the image processing apparatus. FIG. 12 is a functional block diagram of an image processing module that compresses a fixed length of an uncompressed image or the like in a one-dimensional block and packs it together with attribute information. FIG. 13 is an image of data packed with attribute information after fixed-length compression in a one-dimensional block. FIG. 14 is a functional block diagram of an image processing module that compresses an uncompressed image or the like with a one-dimensional block, packs it with attribute information, rotates the packed image data, decodes it, and processes the raster image. . Figure 15: Uncompressed images, etc. are fixed-length compressed in one-dimensional blocks, packed with attribute information, packed image data is decoded and rotated, block data is converted to raster data, and raster data is image processed It is a functional block diagram of an image processing module to perform. FIG. 16 is an explanatory diagram of a rotation access address generation method.

  9 to 16, an image processing unit 9 is means for generating new raster data by performing image processing on raster data. The plotter 10 is means for printing in raster data units. The CPU 11 is an arithmetic processing unit that executes an image processing program. The memory 12 is a means for storing an image processing program and image data. The image rotator 21 is means for rotating the entire square block by rotating the image data collected in the square block in units of pixels within the square block. The separating unit 22 is a unit that separates the fixed-length compressed image from the attribute information. The decompressor 23 is means for returning a fixed-length compressed image to an uncompressed image. The image rotating means 24 is means for rotating the image data in units of pixels. The attribute information rotation means 25 is a means for rotating attribute information in units of pixels. The fixed length compression means 26 is a means for compressing a non-compressed image in a fixed length in units of a plurality of pixels in the line direction, for example, in units of 8 pixels. The packing unit 27 is a unit that combines the fixed-length compressed image and the attribute information into one data. The block-raster converter 28 is means for generating raster data from image data in units of square blocks. The data format converter 29 is means for converting an uncompressed image or the like into packing data of a fixed-length compressed image in units of multiple pixels in the line direction and attribute information. The packing data rotator 30 is means for rotating packing data in units of a plurality of pixels in the line direction into square blocks. The packing data decoder 31 is means for decoding packing data in units of a plurality of pixels in the line direction into raster data. The packing data decoding rotator 32 is means for decoding packing data of a fixed-length compressed image and attribute information in units of a plurality of pixels in the line direction and rotating the image.

  The function and operation of the image processing apparatus according to the second embodiment of the present invention configured as described above will be described. First, the meaning of compressing in units of a plurality of pixels in the line direction will be described. When rotating an image, the fixed-length compression in units of square blocks in the first embodiment is advantageous. When it is desired to handle an image as a raster without rotating it, it is advantageous to perform a fixed length compression in units of a plurality of pixels in the line direction. Since the image rotator reads an image in a two-dimensional block, rotates it, and writes it back to the memory, packing data for a pixels packed in units of multiple pixels in the line direction must be read for a lines in the sub-scanning direction. There is. Data compressed and packed for each of a plurality of pixels in the line direction can be easily extracted with a raster. Since image processing is often performed in units of rasters, it can be said to be an advantageous image format when the decompressed image data is directly input to the image processing unit or output to a plotter that requires input in units of rasters.

  Next, the operation of the image rotator used in the image processing apparatus will be described with reference to FIG. The data size of a fixed-length compressed image of one line (a pixel) is p word. The data size of the attribute information corresponding to it is q word. Therefore, a read from the memory is (p + q) words per access. Since the fixed-length compressed image and the attribute information are encoded a pixel at a time, one block of data can be read out by accessing a times. Data write can also be performed a times. Two DMA controllers are required for reading data before rotation and for writing data after rotation. The image block in FIG. 9 is, for example, an 8-pixel RGB image.

  In the image rotator 21, first, the separation unit 22 separates the fixed length compressed image from the attribute information. The attribute information is rotated by attribute information rotation means 25. The rotated attribute information is made into a square block, and the attribute information is extracted for each of a plurality of pixels in the line direction. The fixed-length compressed image is decoded by the decompressor 23 to be an uncompressed image. The uncompressed image is rotated by the image rotation means 24. The rotated uncompressed image is formed into a square block, and fixed length compression is performed by the fixed length compression means 26 for each of a plurality of pixels in the line direction. The fixed length compressed image and the attribute information are packed by the packing means 27. The format and size of the packed data is equal to the format and size of the packing data before rotation. In this way, the image rotator 21 rotates the fixed-length compressed image and attribute information with a pixel as one block.

  Next, with reference to FIG. 10, the merit of compressing in units of a plurality of pixels in the line direction will be described. Data packed in units of a pixels is read from the memory. When the data packed in units of a pixels is returned to an uncompressed image by the decompressor 23, raster data of a pixels is obtained. The raster data is processed by the image processing unit 9 to obtain processed raster data. When the data packed in units of a pixels is read from the memory and decompressed, the decompressed data is a raster. Therefore, when the image is processed in units of lines without being rotated, it can be directly input to the image processing unit.

  Next, a case where the plotter is output without rotating the image will be described with reference to FIG. Data packed in units of a pixels is read from the memory. When the data packed in units of a pixels is returned to an uncompressed image by the decompressor 23, raster data of a pixels is obtained. This raster data is printed by the plotter 10. It is necessary to input an image with a raster to the plotter. When the data packed in units of a pixels is read from the memory and decompressed, the compressed data is packed by a pixels in the line direction, so that the decompressed image can also be obtained in a raster. The decompressed image data can be output directly to the plotter without being written back to the memory.

  Next, an image processing module that compresses a non-compressed image and the like with a one-dimensional block and packs it together with attribute information will be described with reference to FIG. In FIG. 12, a one-dimensional variable-length code image, a fixed-length code image, and an uncompressed image are shown as examples of input data. One-dimensional means that the compression direction of the data is one-dimensional, and actually a two-dimensional image is handled. The input image is input by some means, converted back to an uncompressed image by the data format converter 29, and stored in the memory 12 together with the attribute information. The input image is compressed by the data format converter 29 for each block of 8 pixels in the raster direction, packed with attribute information, and written to the memory 12 in the format shown in FIG.

  Next, with reference to FIG. 13, a method for packing together with attribute information by compressing a fixed length in units of a plurality of pixels in the line direction will be described. One line of image data is divided into blocks each having a pixels. The block of a pixels is compressed at a fixed length, and it is stored together with its attribute information in a memory. When rotating 90 degrees, in this data format, packing data for a line is read from the memory so as to form a square block. Image data is rotated after decoding the packing data for a line into square blocks. The rotated square block is compressed again to a fixed-length codes every a pixels. Since rotation can be performed in this way, it is not necessary to expand the entire image into an uncompressed image. Although there is an advantage that the amount of memory can be reduced, the number of memory accesses is larger than that in the first embodiment.

  Although the number of memory accesses during rotation is inferior to that of the first embodiment, it is superior to the first embodiment in that raster conversion can be performed without a buffer. A variable-length code can be rotated while being decoded, but cannot be written again as a variable-length code (JPEG) when writing to the memory. It is possible to write back an uncompressed image or a fixed-length compressed image. From the beginning, if it is stored in the memory with a fixed length code that has no correlation between the uncompressed image or adjacent blocks, it can be rotated as it is, but if you convert the format to fixed length code and attribute packing data, There is an advantage that it is easy to perform the rotation process later.

  Next, referring to FIG. 14, an image processing module that compresses a fixed length of an uncompressed image or the like in a one-dimensional block, packs it together with attribute information, rotates and decodes the packed image data, and processes raster data Will be described. 12 is the same as the module of FIG. 12 until the input image is fixed-length compressed for each block of 8 pixels by the data format converter 29, packed together with the attribute information and written to the memory 12. When rotating, the packing data rotator 30 rotates. Since the packing data rotator 30 performs both reading and writing with respect to the memory 12, it is possible to perform both rotation reading and rotation writing. The packed data is decoded by the packing data decoder 31 into raster data. Raster data is processed by the image processing unit 9 to obtain raster data of the processed image.

  In the case of image data that is fixed-length compressed in a one-dimensional block and packed with attribute information, there is no need to have a block-raster conversion line buffer when decoding the packing data and converting it into raster data. However, when an image is rotated, it is impossible to convert it into a raster while performing decoding and rotation, so it is necessary to have a line buffer for block-raster conversion. Alternatively, it is necessary to have a separate module dedicated to rotation and perform rotation in advance via a memory.

  Next, referring to FIG. 15, the uncompressed image is fixed-length compressed with a one-dimensional block, the image data packed together with the attribute information is decoded and rotated, and the block data is converted into raster data. An image processing module for image processing will be described. 12 is the same as the module of FIG. 12 until the input image is fixed-length compressed for each block of 8 pixels by the data format converter 29, packed together with the attribute information and written to the memory 12. The packed data is decoded by the packing data decoding rotator 32 and rotated to form rotated block data. In this case, it is necessary to perform rotation reading, and rotation writing is not possible. Block data is converted into raster data by a block-raster converter 28. Raster data is processed by the image processing unit 9 to obtain raster data of the processed image. Even in the case of image data in which an uncompressed image or the like is fixed-length compressed with a one-dimensional block and packed together with attribute information, it is necessary to have a line buffer for block-raster conversion in order to perform decoding and rotation at the same time.

  Next, a rotation access address generation method will be described with reference to FIG. It is assumed that data is packed every 8 pixels including an image and attribute information. Similar to the uncompressed image of the conventional example, “rotational read” (a method of reading according to the order of the arrangement after rotation) or “rotation write” (a method of writing according to the order of the arrangement after rotation) is possible. Since each block is packed every 8 pixels, at least 8 memory reads are required per block. In the case of rotation read, it is possible to directly rasterize using a block-raster conversion buffer without writing back to the memory after rotation. In the case of a rotating light, direct rasterization is not possible.

  In the case of “rotational read”, a memory read address and a memory write address are generated as follows. First, in order to read the first line block of the A1 block, its head address is obtained. (Total start address) + (Line block size) × (Number of blocks in one line) × (Number of line blocks in one column−8) The head address of the second line block of the A1 block is (head address of the first line block of the A1 block) + (line block size) × (number of blocks of one line). In this way, data is read up to the eighth line block of the A1 block. Rotate the A1 block and write the rotated A1 block to the start address of the rotated memory area. The order of writing is the same as the normal raster order. However, the second line block of the A1 block is (write start address of the first line block of the A1 block) + (line block size) × (number of blocks of one line). The same applies hereinafter. Next, the B1 block is read. This read address is obtained by (A1 block start address) − (line block size) × (number of blocks in one line) × 8. The B1 block is rotated, and the rotated B1 block is written to the next address of the A1 block in the memory area after the rotation. Repeat this. When the rotation of one row is finished, read the A2 block. The read start address of the A2 block is obtained by (read start address of the A1 block) + (square block size). As a square block, the A2 block after the rotation is simply written at the address next to the previous square block.

  In the case of “rotary write”, a memory read address and a memory write address are generated as follows. The reading order is the same as the normal raster order. That is, the D1 block is read first, and thereafter the blocks are read one after another in raster order. However, the head address of the second line block of the D1 block is (head address of the first line block of the D1 block) + (line block size) × (number of blocks of one line). Rotate the D1 block and write the rotated D1 block to the corresponding address in the rotated memory area. The write address of the first line block of the D1 block is obtained by (start address of write area) + (line block size) × (number of blocks in one column−1). The write address of the second line block of the D1 block is (start address of the first line block of the D1 block) + (line block size) × (number of blocks in one column). The write start address of the first line block of the D2 block after rotation is obtained by (write start address of D1 block) + (square block size) × (number of blocks in one column). This is repeated, and when the rotation for 8 lines is completed, the C1 block is read. The write address of the C1 block is obtained by (write start address of D1 block) − (square block size).

The merit of image data that is fixed-length compressed in one-dimensional blocks and packed together with attribute information is shown below.
(1) Since attribute information and image information are packed, the number of memory accesses can be halved.
(2) Since the image itself is compressed at a fixed length, the memory access time is short.
(3) Compared with the data format of the first embodiment, when the data is not rotated, it can be directly converted into raster data without reading the block-raster conversion buffer after reading from the memory.

  As described above, in the second embodiment of the present invention, the image processing apparatus packs image data and attribute information that have been fixed-length compressed in units of 8 pixels in the line direction, stores them in a memory, and converts them into raster data. Since the image processing is performed, the compressed and packed data can be easily extracted by a raster and processed in units of lines.

  The image processing apparatus of the present invention is most suitable as an image processing apparatus that performs high-speed rotation processing on fixed-length compressed image data.

It is a figure of the data format used with the image processing apparatus in Example 1 of this invention. It is a functional block diagram of the rotator used with the image processing apparatus in Example 1 of this invention. It is a figure which shows the method of obtaining raster data from the packed data with the image processing apparatus in Example 1 of this invention. It is a figure which shows the method of obtaining raster data from the packed data, and printing by the image processing apparatus in Example 1 of this invention. 3 is a functional explanatory diagram of a block-raster converter of the image processing apparatus in Embodiment 1 of the present invention. FIG. 2 is a functional block diagram of an image processing module of the image processing apparatus according to Embodiment 1 of the present invention. FIG. It is a functional block diagram of the module of the other image processing of the image processing apparatus in Example 1 of this invention. It is explanatory drawing of the method of rotating an image with the image processing apparatus in Example 1 of this invention. It is a functional block diagram of the rotator used with the image processing apparatus in Example 2 of this invention. It is a figure which shows the case where the image processing apparatus in Example 2 of this invention acquires raster data from the packed data, and performs image processing. It is a figure which shows the case where a plotter is output without rotating with the image processing apparatus in Example 2 of this invention. It is a functional block diagram of the image processing module of the image processing apparatus in Example 2 of the present invention. It is a conceptual diagram of the data which packed the fixed length compression image with the attribute information with the image processing apparatus in Example 2 of this invention. It is a functional block diagram of the image processing module of the image processing apparatus in Example 2 of the present invention. It is a functional block diagram of the image processing module of the image processing apparatus in Example 2 of the present invention. It is explanatory drawing of the method to generate | occur | produce a rotation access address in the image processing apparatus in Example 2 of this invention. It is a figure which shows the packing method and memory access method of the conventional image data. It is a figure which shows the conventional fixed length data packing method. It is a figure which shows the example of the data in which the conventional photograph and a character were mixed. It is a functional block diagram of the conventional image rotator. It is explanatory drawing of the prior art example which rotates an uncompressed image 90 degrees.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Image rotator, 2 ... Separation means, 3 ... Expander, 4 ... Image rotation means, 5 ... Attribute information rotation means, 6 ... Fixed length compression means, 7 ..Packing means, 8 ... block-raster converter, 9 ... image processing unit, 10 ... plotter, 11 ... CPU, 12 ... memory, 13 ... data format converter, 14 ... Decoding rotator, 21 ... Image rotator, 22 ... Separation means, 23 ... Decompression device, 24 ... Image rotation means, 25 ... Attribute information rotation means, 26 ... Fixed length compression means, 27 ... packing means, 28 ... block-raster converter, 29 ... data format converter, 30 ... packing data rotator, 31 ... packing data decoder, 32: Packing data decoding rotator.

Claims (5)

A storage unit for storing image data ;
Means for blocking and fixing the image data in units of square areas; means for blocking the attribute information of the image data in units of the same block; and the image data compressed in fixed lengths in blocks of the square area units And a two-dimensional block data format conversion unit comprising: packing attribute information for each block and storing the attribute information in the storage unit;
Means for blocking the image data in units of pixels in the line direction and compressing the fixed length; means for blocking attribute information of the image data in units of blocks in the same line direction; and blocks in units of pixels in the line direction A one-dimensional block data format conversion unit comprising: means for packing the image data and the attribute information that have been fixed-length-compressed in block by block and storing them in the storage unit;
The block with the direction of each square area unit block fixed in a block group constituted by a plurality of square area unit blocks in the main scanning direction and the sub-scanning direction stored in the storage unit by the two-dimensional block data format conversion unit An image rotation unit that rotates the group by an integral multiple of 90 °;
An image processing apparatus comprising: a decompression unit that decodes packing data stored in the storage unit by the one-dimensional block data format conversion unit in units of blocks in a line direction and converts the data into raster data . For the data in which fixed-length compressed image and attribute information are packed for each square area unit block, means for generating an address rotated 90 ° for each square area unit block unit, and square area unit by one memory read access according to the rotated address The means for reading a block from the storage unit, an image rotator for rotating the read block data, and a unit for writing the rotated block data to the storage unit in raster order. The image processing apparatus according to 1. Means for reading out data from the storage unit in raster order in units of square area unit blocks for data in which fixed-length compressed images and attribute information are packed for each square area unit block, and an image rotator for rotating the read block data And means for generating an address rotated by 90 ° in block units, and means for writing the rotated square area unit block into the storage unit by one memory write access according to the rotated address. Item 6. The image processing apparatus according to Item 1 . The image processing apparatus according to claim 1, wherein the image data is uncompressed image data . The image processing apparatus according to claim 1 , wherein the image data is variable length compressed image data or fixed length compressed image data .

Images