JP2011103696A - Image compressing method and image compressing apparatus - Google Patents

Abstract

The present invention provides an image compression method and an image compression apparatus capable of providing redundancy and increasing the compression efficiency in encoding processing.
In an image compression method for performing compression processing on an image including a recognition target, an image including the recognition target is input, and a predetermined filter process is performed on the input image to obtain an intermediate value in the filter. The acquired intermediate value is compared with the current pixel value of interest, and if the difference between the two is equal to or greater than a predetermined threshold value, no processing is performed on the pixel of interest. The process of replacing the target pixel value of the target pixel value with the acquired intermediate value is repeatedly performed until the target pixel covers all the pixels, and smoothing processing is performed. The smoothed image is variable according to the appearance frequency. Perform long coding.
[Selection] Figure 10

Description

  Embodiments described herein relate generally to an image compression method and an image compression apparatus that perform compression processing on an image including a recognition target in, for example, a mail processing apparatus or a face collation apparatus.

  For example, in a mail processing apparatus, there is a problem in the transmission time of an image having a resolution necessary for character recognition or the like due to processing time restrictions. Also, in the face collation device, the collated face image is recorded as a log for analysis, and there is a problem of the log image extraction time at the customer site. In order to solve these problems, a method for reducing the transmission cost by compressing an image or a method for reversibly compressing an image is known.

  The method of reducing the transmission cost by compressing the image is to calculate the adjacent pixel difference value in the encoding of the still image data by the lossless method, and assign the variable length code according to the appearance frequency, This is an image coding method capable of performing variable length coding with a good compression rate.

  The method of reversibly compressing an image is a reversible and efficient method by calculating positive or negative binary information and absolute value of the difference value of luminance information of adjacent pixels, and performing bitmap development for each pixel and encoding. This is a multi-value information compression method capable of compression.

  However, in the conventional method, encoding with a good compression rate can be performed in lossless compression, but the compression rate is limited because of lossless compression. Further, in general irreversible compression such as so-called JPEG, a feature to be recognized is lost, which affects recognition.

JP-A-9-116765 JP 2000-156861 A

  The problem to be solved by the present invention is to provide an image compression method and an image compression apparatus capable of providing redundancy and improving the compression efficiency in the encoding process.

  An image compression method according to an embodiment is an image compression method for performing compression processing on an image including a recognition target, an image input step for inputting an image including the recognition target, and an image input by the image input step. Performs smoothing processing, obtains an intermediate value in the filter by performing a predetermined filtering process on the image input in the image input step, and compares the acquired intermediate value with the current target pixel value If the difference between the two is equal to or greater than a predetermined threshold value, no processing is performed on the target pixel. If the difference is less than the threshold value, the target pixel is replaced with the acquired intermediate value. A smoothing step that is repeated until all pixels are targeted, and a difference value between adjacent pixels for the image smoothed by the smoothing step is calculated. A difference value calculation step, and a coding step of performing variable-length coding in accordance with the appearance frequency to the difference value calculated by the difference value calculation step.

1 is a block diagram schematically showing the configuration of an image compression apparatus according to a first embodiment. Explanatory drawing which shows the processing content of a smoothing process part. The flowchart which shows the process sequence of a smoothing process part. FIG. 5 is a block diagram schematically showing the configuration of an image compression apparatus according to a second embodiment. Explanatory drawing showing the processing content of a feature extraction part. Explanatory drawing showing the processing content of a gradation conversion part. FIG. 9 is a block diagram schematically showing the configuration of an image compression apparatus according to a third embodiment. The flowchart which shows the process sequence of a difference object discrimination | determination part. Explanatory drawing showing the processing content of a difference value calculation process part. FIG. 10 is a block diagram schematically showing the configuration of an image compression apparatus according to a fourth embodiment. The flowchart which shows the process sequence of a smoothing difference process part. FIG. 10 is a block diagram schematically showing the configuration of an image compression apparatus according to a fifth embodiment. Explanatory drawing showing the processing content of an outlier confirmation part.

Hereinafter, embodiments will be described with reference to the drawings.
First, the first embodiment will be described.
FIG. 1 schematically shows the configuration of an image compression apparatus according to the first embodiment. The image compression apparatus according to the first embodiment includes an image input unit 101 as an image input unit, a smoothing processing unit 102 as a smoothing unit, an encoding unit 103 as an encoding unit, and a data output unit 104. It is configured.

  The image input unit 101 includes, for example, a television camera or an image scanner, collects an image including a recognition target, and inputs the collected image to the smoothing processing unit 102. Specifically, for example, it is collected and inputted as a digital grayscale image having 640 pixels in the horizontal direction and 480 pixels in the vertical direction.

The smoothing processing unit 102 performs smoothing processing on the image input by the image input unit 101 and outputs the processing result to the encoding unit 103. Specifically, for example, as shown in FIG. 2, 3 × 3 median filter processing is performed.
[Equation 1]
｜ Median (f (x + i, y + i))-f (x, y) |> = Th f (x, y) = f (x, y)
｜ Median (f (x + i, y + i))-f (x, y) | <Th f (x, y) = Median (f (x + i, y + i))
Hereinafter, a series of flow of the filtering process will be described with reference to a flowchart shown in FIG. First, an intermediate value in the filter (3 × 3) is acquired by a 3 × 3 median filter as shown in FIG. 2 (step S1). Next, the acquired intermediate value is compared with the current pixel value of interest (step S2), and if the difference between the two is equal to or greater than a predetermined threshold value, no processing is performed on the pixel of interest, and is less than the threshold value. In this case, the current pixel value of interest is replaced with the intermediate value acquired in step S1 (step S3). The processing result is obtained by repeating these processes until the target pixel covers all the pixels. Smoothing processing is realized by the filter processing as described above.

  The encoding unit 103 performs variable length encoding processing on the image obtained from the smoothing processing unit 102, and outputs the processing result to the data output unit 104. Specifically, first, the appearance frequency for each data value obtained for the image obtained from the smoothing processing unit 102 is calculated. If the appearance frequency for each data value is known, it is possible to assign a Huffman code to each data value by assigning a code with a smaller number of bits to a data value with a higher appearance frequency, and variable length based on Huffman code theory Encoding is performed. Although the Huffman encoding method is used in the present embodiment, the Huffman encoding method is not necessarily used as the encoding method, and for example, run length encoding may be used.

  The data output unit 104 outputs the encoded image obtained from the encoding unit 103 as a compressed image.

  As described above, according to the first embodiment, redundancy is provided by performing smoothing processing suitable for compression while retaining important image information (for example, a steep luminance gradient). The compression efficiency in the conversion process can be increased.

Next, a second embodiment will be described.
FIG. 4 schematically shows a configuration of an image compression apparatus according to the second embodiment. An image compression apparatus according to the second embodiment includes an image input unit 201 as an image input unit, a feature extraction unit 202 as a feature extraction unit, a gradation conversion processing unit 203 as a gradation conversion unit, and an encoding unit. It comprises an encoding unit 204 and a data output unit 205.

  Since the image input unit 201, the encoding unit 204, and the data output unit 205 are the same as the image input unit 101, the encoding unit 103, and the data output unit 104 in the first embodiment described above, description thereof is omitted. .

  The feature extraction unit 202 extracts a luminance feature to be recognized from the image input by the image input unit 201 and outputs it to the gradation conversion processing unit 203. Specifically, for example, when the recognition target is a letter, two features, a character area and a letter area (background area), are obtained as features. Therefore, a luminance histogram is obtained as shown in FIG. 5, and two maximum values (Max1, Max2) are extracted. Since the extracted luminance value near the maximum value represents a character area and a letter area, threshold values (Th1, Th2) used by the gradation conversion processing unit 203 are obtained from the maximum value. As a method of obtaining the threshold value, a value that is a fixed value away from the maximum value may be simply used as the threshold value (Th1, Th2), or a value in which the cumulative value of the luminance histogram becomes constant may be used as the threshold value (Th1, Th2). .

The gradation conversion processing unit 203 performs gradation conversion processing for assigning a fixed value based on the threshold values (Th1, Th2) obtained from the feature extraction unit 202, and outputs the processing result to the encoding unit 204. Specifically, for example, when the recognition target is a letter, it can be converted by performing the calculation represented by the following formula 2 using the threshold values (Th1, Th2) obtained from the feature extraction unit 202. An explanatory diagram of this gradation conversion is shown in FIG.
[Equation 2]
f (x, y) <= Th1 f ′ (x, y) = Th1
f (x, y)> = Th2 f ′ (x, y) = Th2
As described above, according to the second embodiment, it is possible to provide redundancy by performing gradation conversion that does not affect recognition, and to increase the compression efficiency in the encoding process.

Next, a third embodiment will be described.
FIG. 7 schematically shows a configuration of an image compression apparatus according to the third embodiment. An image compression apparatus according to the third embodiment includes an image input unit 301 as an image input unit, a face region detection unit 302 as a face region detection unit, a difference target determination unit 303 as a difference target determination unit, and a difference value calculation unit. As a difference value calculation processing unit 304, an encoding unit 305 as an encoding unit, and a data output unit 306.

  Note that the encoding unit 305 and the data output unit 306 are the same as the encoding unit 103 and the data output unit 104 in the first embodiment described above, and thus description thereof is omitted.

  The image input unit 301 is configured by, for example, a television camera, collects a plurality of frames of images including faces for a person who performs face matching, and inputs the collected images to the face area detection unit 302. Specifically, for example, it is collected and inputted as a digital grayscale image having 512 pixels in the horizontal direction and 512 pixels in the vertical direction.

The face area detection unit 302 cuts out a face pattern (face area) from the image input by the image input unit 301. This is described in, for example, literature (Kazuhiro Fukui, Osamu Yamaguchi: “Face Feature Point Extraction by Combination of Shape Extraction and Pattern Matching”, IEICE Transactions (D), vol. J80-DH, No. 8, By using the method described in pp2170-2177 (1997), it is possible to detect a face area with high accuracy.
In this case, by obtaining a correlation value while moving a template prepared in advance in the image, a place having the highest correlation value is set as a face region. In addition, a face extraction method using an eigenspace method or a subspace method may be used.

The difference target determination unit 303 determines a difference target region from the face region detected by the face region detection unit 302 and outputs the difference target region to the difference value calculation processing unit 304. Specifically, for example, as in the flowchart shown in FIG. 8, the ratio of the face area detected by the face area detection unit 302 to the entire image input by the image input unit 301 is obtained (step S11), and obtained. If the ratio is equal to or greater than a predetermined threshold, the difference target area is set as a face area, and the face area is aligned (step S12). .
Note that the alignment of the face area can be realized, for example, by performing the alignment using the center of gravity of the face area acquired by the face area detection unit 302.

The difference value calculation processing unit 304 calculates the difference value of the corresponding pixels between consecutive frames as shown in the following equation 3 for the image determined and processed by the difference target determination unit 303, and sends it to the encoding unit 305. Output.
[Equation 3]
f _t (x, y) = _ft (x, y) −f _{t + 1} (x, y)
That is, as shown in FIG. 9, the difference value calculation process can be efficiently performed without depending on the data by switching the difference target area between the face area and the background area according to the characteristics of the input image. FIG. 9A shows a case where the difference target area is the background area, and the difference between the background images (shaded portions) is taken. FIG. 9B shows a case where the difference target area is a face area, and the difference between face images (shaded portions) is taken.

  As described above, according to the third embodiment, it is possible to provide redundancy by performing appropriate difference value calculation processing using the features of the face image, and to increase the compression efficiency in the encoding processing. .

Next, a fourth embodiment will be described.
FIG. 10 schematically shows a configuration of an image compression apparatus according to the fourth embodiment. An image compression apparatus according to the fourth embodiment is a modification of the above-described first embodiment, and includes an image input unit 401 as an image input unit, a smoothing unit, and a smoothed difference processing unit as a difference value calculation unit. 402, an encoding unit 403 as an encoding unit, and a data output unit 404.

  Note that the image input unit 401, the encoding unit 403, and the data output unit 404 are the same as the image input unit 101, the encoding unit 103, and the data output unit 104 in the above-described first embodiment, and thus description thereof is omitted. .

  The smoothing difference processing unit 402 performs smoothing processing on the input image data first, then performs difference value calculation processing between adjacent pixels, and outputs the processing result to the encoding unit 403.

  Hereinafter, the processing of the smoothing difference processing unit 402 will be described with reference to the flowchart shown in FIG. First, an intermediate value in the filter (3 × 3) is acquired by a 3 × 3 median filter as shown in FIG. 2 (step S21). Next, the acquired intermediate value is compared with the current pixel value of interest (step S22). In this case, the current pixel value of interest is replaced with the intermediate value acquired in step S21 (step S23).

  Next, a difference value calculation process between adjacent pixels is performed (step S24). Next, it is checked whether or not the search for all pixels has been completed (step S25). If the search for all pixels has not been completed, the target pixel is updated (step S26), and then the process returns to step S21. The same processing as above is repeated. Then, the process ends when the search for all the pixels is completed.

Specifically, the smoothing process can be realized by a method similar to that of the smoothing processing unit 102 in the first embodiment. Also, the difference value calculation process between adjacent pixels can be calculated using the following equation (4).
[Equation 4]
f (x, y) = f (x, y) −f (x−1, y)
As described above, according to the fourth embodiment, smoothing is performed while retaining important image information (for example, a steep luminance gradient), and redundancy suitable for compression is provided in the difference value calculation process. By doing so, the compression efficiency in the encoding process can be increased.

Next, a fifth embodiment will be described.
FIG. 12 schematically shows a configuration of an image compression apparatus according to the fifth embodiment. An image compression apparatus according to the fifth embodiment is a modification of the above-described second embodiment, and includes an image input unit 501 as an image input unit, a face area detection unit 502 as a face area detection unit, and an outlier calculation. An outlier confirmation unit 503 as a means, a gradation conversion processing unit 504 as a gradation conversion unit, an encoding unit 505 as an encoding unit, and a data output unit 506 are configured.

  Note that the image input unit 501, the gradation conversion processing unit 504, the encoding unit 505, and the data output unit 506 are the image input unit 201, gradation conversion processing unit 203, encoding unit 204, and encoding unit 204 in the second embodiment described above. The data output unit 205 is the same, and the face area detection unit 502 is the same as the face area detection unit 302 in the third embodiment described above, and thus the description thereof is omitted.

  The outlier confirmation unit 503 obtains an outlier from the luminance feature of the face area detected by the face region detection unit 502 and the luminance feature of the input image, and outputs the outlier to the gradation conversion processing unit 504. For example, as shown in FIG. 13, a face area histogram is obtained, and a maximum value (Max) and a minimum value (Min) are obtained from the obtained face area histogram. Values that are apart from the determined maximum and minimum values (Max, Min) by a certain value are set as threshold values (Th1, Th2).

  As described above, according to the fifth embodiment, even for a face image having a complicated luminance feature, redundancy is provided by performing gradation conversion that does not affect recognition. The compression efficiency can be increased.

As described above, according to the above-described embodiment, redundancy is provided by performing smoothing processing suitable for compression while retaining important image information (for example, a steep luminance gradient). The compression efficiency can be increased.
Further, by performing gradation conversion that does not affect recognition, redundancy can be provided, and compression efficiency in the encoding process can be increased.

  Furthermore, redundancy can be provided by performing an appropriate difference value calculation process using the features of the face image, and the compression efficiency in the encoding process can be increased.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the scope of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

  101, 201, 301, 401, 501 ... image input unit (image input unit), 102 ... smoothing processing unit (smoothing unit), 103, 204, 305, 403, 505 ... encoding unit (encoding unit), 104, 205, 306, 404, 506 ... data output unit, 202 ... feature extraction unit (feature extraction means), 203, 504 ... gradation conversion processing unit (gradation conversion means), 302, 502 ... face area detection unit ( Facial area detection means), 303 ... Difference object determination section (difference object determination means), 304 ... Difference value calculation processing section (difference value calculation means), 402 ... Smoothing difference processing section (smoothing means, difference value calculation means) 503: Outlier confirmation unit (outlier calculator).

Claims (2)

In an image compression method for performing compression processing on an image including a recognition target,
An image input step of inputting an image including the recognition target;
A smoothing process is performed on the image input in the image input step, and an intermediate value in the filter is acquired by performing a predetermined filter process on the image input in the image input step. The intermediate value is compared with the current target pixel value, and if the difference between the two is equal to or greater than a predetermined threshold value, no processing is performed on the target pixel. A smoothing step of repeatedly performing the process of replacing with the acquired intermediate value until the target pixel covers all the pixels;
A difference value calculating step for calculating a difference value between adjacent pixels with respect to the image smoothed by the smoothing step;
An encoding step for performing variable length encoding on the difference value calculated by the difference value calculating step according to its appearance frequency;
An image compression method comprising: In an image compression apparatus that performs compression processing on an image including a recognition target,
Image input means for inputting an image including the recognition target;
A smoothing process is performed on the image input by the image input means, and an intermediate value in the filter is acquired by performing a predetermined filter process on the image input by the image input means. The intermediate value is compared with the current target pixel value, and if the difference between the two is equal to or greater than a predetermined threshold value, no processing is performed on the target pixel. Smoothing means for repeatedly performing the process of replacing with the acquired intermediate value until the target pixel covers all pixels;
A difference value calculating means for calculating a difference value between adjacent pixels with respect to the image smoothed by the smoothing means;
Encoding means for performing variable length encoding on the difference value calculated by the difference value calculating means according to its appearance frequency;
An image compression apparatus comprising:

Images