US20100278389A1 - Method for image recombination of a plurality of images and image identification and system for image acquiring and identification - Google Patents

Abstract

The present invention provides a method for image recombination of a plurality of images and image identification and a system for image acquiring and identification. Features with respect to the plurality of images are recombined and enhanced so as to form a recombined image. After that, the recombined image is processed to emphasize the features of the recombined image so that the recombined image is capable of being identified easily. Furthermore, the present provides a system to perform the foregoing method, whereby reducing unidentified problems caused due to low quality image of the monitoring system.

Description

FIELD OF THE INVENTION
• The present invention generally relates to an image identification technology and, more particularly, to a method for image recombination of a plurality of images and image identification and a system for image acquiring and identification, wherein features with respect to the plurality of images are recombined so as to identify the content of an image.
BACKGROUND OF THE INVENTION
• There are many people who get killed in traffic accidents. Theft and burglary using cars/motorcycles have been repeatedly reported. These may be attributed to poor image identification of license plates because of poor monitoring systems. Such monitoring systems are mostly problematic because of poor resolution (320×240 Pixels) and slant angles of the image acquiring units to cause blur or imcomplete images that cannot be recognized so that the criminals can be at large.
• Conventionally, in Taiwan Patent No. 197752, a CCD camera and an image acquiring unit are used to acquire a car image in the car lane and the car image is then read by an image reading unit. Then, a logarithmic greyscale operation unit is used to calculate the logarithmic greyscale of each pixel in the car image. The image corresponding to the logarithmic greyscales is decomposed by a wavelet decomposition unit into rough images, horizontally differentiated images, vertically differentiated images and diagonally differentiated images. An image binarization unit converts the logarithmic greyscale of each pixel in the horizontally differentiated images from real numbers intobinary digits 0 and 1. A rough image dividing unit determines a region with the highest sum of binary digits within the whole car image according a pre-set license plate size and thus the region is initially referred to as a license plate region. Then, a license plate slantness correction unit corrects the slantness of the image corresponding to the license plate region. Finally, a fine image dividing unit removes the part that does not correspond to the license plate from the rough license plate region.
• Moreover, in Taiwan Patent Pub. No. I286027, an integrated plurality of lane free flow vehicle enforcement system is disclosed, wherein a portal framed equipment is established at the image enforcement point. The car lane is physically divided so that image enforcement can be realized with respect to various cars even though the system slows the cars to pass by the image enforcement point at a normal speed and to change lanes freely.
• Moreover, in Taiwan Patent Appl. No. 200802137, a serial license plate identification system is disclosed, using a license plate character region detection module to receive an image and determine each approximate license plate range in the image. Sequences of serial identical pixels in each approximate license plate range are obtained. The sequences of serial identical pixels are erased, filtered, and connected to blocks so as to obtain the image with respect to the license plate character region in each approximate license plate range and output verified image with respect to the license plate character region after verification. Then, the verified image with respect to the license plate character region is transmitted to the a license plate character dividing and identification module to acquire all the independent character images and thus all the license plate character information after the independent character images are identified.
• Moreover, Taiwan Patent No. 221193 discloses a license plate identification and monitoring apparatus used in a parking area. When a car passes by a predetermined image acquiring spot, the host is informed to enable the duplex image acquiring device to control the camera device to acquire the car license plate image, which is then processed by the an identification process to identify the characters on the license plate for car management, stolen car seeking and prevention in the parking area.
• Taiwan Patent No. 226454 discloses a license plate identification method, wherein the logic relation and character strokes are used to determine the correct license plate location in the digital image. Then, ternarized difference and fuzzy inference are used to acquire the outlines of the characters on the license plate. Adaptive binarization method is used to divide the boundaries of each character. Finally, the identification result can be obtained by a feature fused median calculation using a neural network.
• Moreover, Taiwan Patent No. 191905 discloses a automatic mobile license plate identification system, which comprises an image acquiring device and an image processing device that can be installed in a car to perform automatic identification on a static or moving car being monitored. The image acquiring device is capable of acquiring the image of the license plate and transmitting the image into the image processing device. The image processing device performs a precise acquiring process on the license plate characters based on fuzzy inference and performs a character identification process on the characters using character structure analysis. Therefore, identification errors due to license plate contamination, bending, character contamination or deflexion can be prevented.
• Taiwan Patent No. 123259 discloses a license plate number identification apparatus installed at a spot where cars pass by so as to automatically identify the license plate number of a car. The license plate number identification apparatus uses an image acquiring device capable of acquiring an image containing the license plate and an image processing unit capable of checking the digital image according to features of the license plate number to find the license plate location, specify the range of characters, divide the characters to achieve feature identification of each characters.
• Moreover, U.S. Pat. No. 4,817,166 discloses a method for reading a license plate by acquiring the boundary features such as length, height, and width of the characters on the license plate. With such information regarding the character features, geometric features of the characters such as the locations and shapes of convex hulls, turns and holes are analyzed. Finally, the structure of each character on the license plate is analyzed according to the results of the aforegoing analysis.
• Moreover, U.S. Pat. No. 6,553,131 discloses an identification technology using a smart image acquiring device to perform license plate identification. A processor is installed inside the image acquiring device to perform license plate information identification. In this technology, image identification is implemented by determining a base line according to the brightness and location of the license plate image and a blur region. The image having a base line is then processed by projection to obtain the location of each character on the license plate. A statistic-based method is used so that each character is provided with a confidence index. Finally, character information on the license plate is determined according to the confidence index.
• Moreover, U.S. Pat. No. 5,425,108 discloses a license plate image identification technology, wherein the acquired license plate image is processed by fuzzy interfere and the features of the license plate image is identified by structure analysis using a neural network features.
• Moreover, U.S. Pat. No. 6,473,517 discloses a license plate identification technology, wherein the license plate image is identified by character segmentation. In this technology, the license plate image is divided into a plurality of regions to be converted into possible character regions (or suspected character regions). Then, the possible character regions are identified to obtain a confidence index for image identification based thereon.
• U.S. Pat. No. 5,081,685 discloses a license plate identification technology, wherein image intensity information is used to identify the characters on the license plate. In this technology, the characters are separated from the background on the license plate so as to obtain the outlines of the characters by a tracking process.
SUMMARY OF THE INVENTION
• The present invention provides a method for image recombination of a plurality of images and image identification and a system for image acquiring and identification, wherein a plurality of images corresponding to a specific target are recombined to compensate the incomplete information of each image to form a recombined image with enhanced image identification.
• The present invention provides a method for image recombination of a plurality of images and image identification and a system for image acquiring and identification, wherein a recombined image formed by recombining a plurality of images is identified with a plurality of similarity indexes according to the provided information in a database. The possible results are obtained according to the similarity indexes for the user to choose from with enhanced identification speed and precision.
• The present invention provides a method for image recombination of a plurality of images and image identification and a system for image acquiring and identification, which is useful in identification on the identification mark of a carrier. With enhanced character features of the identification mark, multi-angle license plate identification technology can be used to help the user to identify cars that are suspected to cause accidents.
• In one embodiment, the present invention provides a method for image recombination of a plurality of images, comprising steps of: acquiring a plurality of images; determining a region of interest in an image from the plurality of images and acquiring features in the region of interest; acquiring from other regions a feature region corresponding to the region of interest according to the acquired features; and performing an image recombination process to form a recombined image according to a plurality of feature regions and the region of interest.
• In another embodiment, the present invention provides a method for image identification, comprising steps of: acquiring a plurality of images; determining a region of interest in an image from the plurality of images and acquiring features in the region of interest; acquiring from other regions a feature region corresponding to the region of interest according to the acquired features; and performing an image recombination process to form a recombined image according to a plurality of feature regions and the region of interest; and identifying the recombined image.
• In another embodiment, the present invention provides a system for image acquiring and identification, comprising: an image input unit capable of providing a plurality of images; an image processing unit coupled to the image input unit, the image processing unit further comprising: a feature acquiring unit capable of acquiring features in a region of interest on an image from the plurality of images and acquiring from other regions a feature region corresponding to the region of interest according to the acquired features in other regions; a recombination unit capable of performing an image recombination process according to the plurality of feature regions and the region of interest to form a recombined image.
BRIEF DESCRIPTION OF THE DRAWINGS
• The objects and spirits of the embodiments of the present invention will be readily understood by the accompanying drawings and detailed descriptions, wherein:
• FIG. 1 is a flowchart of a method for image recombination of a plurality of images according to one embodiment of the present invention;
• FIG. 2A andFIG. 2B are schematic diagrams showing a car that is moving;
• FIG. 3A andFIG. 3B are schematic diagrams showing the acquired images of a car at different locations;
• FIG. 3C is a schematic diagram showing the angular relation with respect to the feature region and the region of interest;
• FIG. 4A andFIG. 4B are schematic diagrams of a plurality of images and a recombined image respectively;
• FIG. 5 is a flowchart of a method for image recombination of a plurality of images according to another embodiment of the present invention;
• FIG. 6A andFIG. 6B are schematic diagrams of a plurality of images and a recombined image respectively;
• FIG. 7A andFIG. 7B show the greyscale value with respect to the location before and after histogram equalization respectively;
• FIG. 8 is a flowchart of a method for image identification according to the present invention;
• FIG. 9A toFIG. 9D are schematic diagrams showing the formation of a sample image;
• FIG. 10A is a schematic diagram of a recombined image and a feature image thereof;
• FIG. 10B is a schematic diagram of a feature image;
• FIG. 11 is a table for sorting the comparison results of an identification mark according to the present invention; and
• FIG. 12 is a schematic diagram of a system for image acquiring and identification according to the present invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
• The present invention can be exemplified but not limited by various embodiments as described hereinafter.
• Please refer toFIG. 1, which is a flowchart of a method for image recombination of a plurality of images according to one embodiment of the present invention. In the present embodiment, themethod2 starts withstep200 to acquiring a plurality of images. In the present step, the plurality of images can be acquired in many ways, for example, inputting a plurality of images or using a camera, a CCD or a CMOS image acquiring unit, but not limited thereto, to acquire the plurality of images at different timings or using an image acquiring device to acquire images with sequential reation from continuous images by a camera. InFIG. 2A, when acar90 is making a turn, images of thecar90 are acquired at different timings by animage acquiring unit25 to acquire a plurality of images. InFIG. 2B, images at different locations are acquired when thecar90 is moving. For example, the image atlocation91 is as shown inFIG. 3A, while the image atlocation92 is as shown inFIG. 3B. The number of the images is based on the practical demand, and is thus not limited. Instep200, the images are images of a car, but not limited thereto.
• Referring toFIG. 1, after the images are acquired,step201 is performed to load the plurality of images. Then, instep212, the images are stored in a storage medium, such as a hard disk or memory. Then step202 is performed to choose from the plurality of images an image as a standard image. The chosen image can be the clearest one to be the standard image. In the embodiment, the standard image is as shown inFIG. 3B. Then, step203 is performed to determine a region of interest (ROI) in the standard image. The region of interest covers the license plate image of a car, as shown inregion93 inFIG. 3B. Then, step204 is performed to acquire the image features in the region of interest. In the present embodiment, the image features include contrast or greyscale value of the outlines of the word or digit on the license plate image in the feature region.
• Then, step205 is performed to acquire from other regions a feature region corresponding to the ROI according to the features acquired instep204. The feature region is acquired manually or automatically with the use of software. The feature region is determined according to the object to be identified. In the present embodiment, an identification mark of a car is used as an example. The feature region refers to a region corresponding to the identification mark. Step205 comprises two steps as described herein. Firstly, the image stored instep212 is read. Then, the features acquired instep204 are loaded to perform a feature searching process on the loaded images. For example, afeature region94 can be searched from the image inFIG. 3A according to features acquired inregion91 inFIG. 3B. Instep205, in addition to acquiring the feature region, the angular relation and the scale relation between the feature region and the region of interest are also determined. For example, inFIG. 3C, after thefeature region94 is acquired,step25 further acquires alocation940 in thefeature region94 corresponding to aspot930 in the region of interest by geometric matching. Then, according to thelocation940, the angular relation based on the coordinates of thefeature region94 is obtained. Furthermore, normalization can be achieved based on the angular relation (θ) and scale relation between thefeature region94 and the region ofinterest93.
• Referring toFIG. 1, instep206, the feature region in other images is acquired repeatedly until the number of processed car images reaches the number of plurality of images. Afterwards,step207 is performed to normalize the acquired feature regions. The step of normalization is performed according to the angular relation and the scale relation instep205 so that the size of each feature region is adjusted to be identical to the size of the region of interest or that the sizes of the feature region and the region of interest are adjusted to a specific scale. Since there are a plurality of images acquired instep200, the size of the target (i.e., a car in the present embodiment) may vary, as shown inFIG. 3A andFIG. 3B, due to different view angles and distances. As a result, thefeature regions93 and94 acquired instep205 may differ in size. Therefore,step207 is performed to adjust the size of each feature region to be identical. In the present embodiment, the size is 130×130 pixels.
• Then, step208 performs an inversing operation on the pixels in the plurality of feature regions and the pixels in the region of interest respectively. The inversing operation is achieved so that the bright turns dark and the dark turns bright. The sensitivity of human eyes is saturated in a bright environment, which leads to failure in identifying the detailed structure in the bright region. Therefore, it is easier to identify the images after the inversing operation is performed to turn the bright into dark. Afterwards,step209 is performed to form a recombined image by summing the pixels in the plurality of feature regions and the pixels in the region of interest. Since the feature region and the region of interest have been normalized to have the same image size, the greyscale values of the pixels can be summed. Instep210, an inversing operation is performed on the recombined image. Then, step211 is performed to enhance the recombined image. The image is enhanced to improve the contrast and the brightness. Referring toFIG. 4A andFIG. 4B,FIG. 4A shows unidentified feature regions on a plurality of images (3 images in the present embodiment) before recombination whileFIG. 4B is a schematic diagram of a recombined image by the method shown inFIG. 1. In other words, the 3 images in FIG.4A are recombined to obtain a clear image.
• Please refer toFIG. 5, which is a flowchart of a method for image recombination of a plurality of images according to another embodiment of the present invention. In the present embodiment, step300 to step307 in themethod3 are similar to step200 to step207 inFIG. 1, and descriptions thereof are thus not presented herein. Themethod3 in the present embodiment is different fromFIG. 1 in that the operation for obtaining the recombined image is different. After normalization instep307,step308 is performed to average the feature regions in the plurality of images to obtain a recombined image. Each pixel in the feature region and region of interest is summed and averaged to form the recombined image. Then, instep309, a histogram equalization process is performed on the recombined image. Th histeogram equalization process is aimed at enhancing the contrast of the recombined image. For example,FIG. 6A shows unidentified feature regions on a plurality of images (3 images in the present embodiment) before recombination, wherein the features on each image are blur and unclear. Afterstep308 and step309 Are performed, a clear image is formed as shown inFIG. 6B.FIG. 7A andFIG. 7B show the greyscale value with respect to the location before and after histogram equalization respectively. It is observed that the contrast difference d is small (FIG. 7A) before the histogram equalization process instep309, while the contrast difference D is large (FIG. 7B) after the histogram equalization process instep309. A larger contrast difference D (FIG. 7B) is helpful for image identification. Referring toFIG. 5,step310 is performed to enhance the features in the recombined image for image identification.
• Please refer toFIG. 8, which is a flowchart of a method for image identification according to the present invention. The flowchart of the method for image identification can be achieved using the recombined image formed inFIG. 1 orFIG. 5 to obtain identification results with respect to the recombined image. In other words, the recombined image formed inFIG. 1 orFIG. 5 is used to obtain information according to the comparison with the plurality of sample images. Themethod4 starts withstep40 to provide a database. The database provides a plurality of standard sample images. Please refer toFIG. 9A, which is a schematic diagram of a sample image. The size of thesample image5 is determined according to the user's demand, for example, 130×130 pixels, but not limited thereto. Astandard image region50 is formed on the pixel in thesample image5. Thestandard image region50 comprises a plurality ofpixels500 and501 to form a character, a digit, a word or a pattern as represented by the sample image. Referring toFIG. 9B, the present embodiment is exemplified by a digit “1”. In thesample image5, eachpixel500 and501 is given a proper greyscale value to form astandard image region50, which draws the outline of thedigit 1. Then, in thestandard image region50, specific pixels501 (pixels with oblique lines) are given a specific weight value. The greyscale value and the weight value are determined according to the user's demand. That is, each weight value may be different or identical. In the present embodiment, the weight value is positive. In thestandard image region50, the greyscale value and the weight value for eachpixel500 and501 are combined as the first feature value.
• In the sample image, thenon-standard image region51 is provided as shown inFIG. 9C. Thenon-standard image region51 represents the content that thestandard image region50 is taken for. For example, digit “1” is often taken for letter “I” or “L” or even letter “E. Therefore, locations forpixels510 possibly mis-identified (pixels with dots) are given proper greyscale values and weight values as the second feature values corresponding topixels510. In the present embodiment, locations for thepixels510 in thenon-standard image region51 are determined according to the easily mis-identified character, digit or word in thestandard image region50, which is not restricted. The greyscale values and weight values are determined according to practical demand. In the present embodiment, the weight values in thenon-standard image region51 are negative.
• As shown inFIG. 9D, which is a schematic diagram showing anothersample image5aprovided according todigit 0, thesample image5aalso comprises a standard image region and a non-standard image region. The pattern constructed by the pixels in the standard image region draws the outline of a digit “0”. Similarly, the pattern constructed by the pixels in the non-standard image region denotes a word that digit “0” is taken for. For example, digit “0” is often taken for letter “Q” or digit “8”. Steps221 and222 can be performed using image processing software exemplified by, but not limited to, MS Paint. The sample images, such as 0 to 9, A to Z and a to z, are stored in the database. Then, in step224, the identification result is observed after a plurality times of training.
• Referring toFIG. 8,step41 is performed to acquire a feature image from the recombined image. For example, inFIG. 10A, the region in the recombined image95 (the image inFIG. 6B) corresponding to each unidentified word is the feature image. Instep41, the acquiredfeature image96 is the first character in the identification information. Then, step42 performs a calculation on a third feature value of each pixel in the feature image and the first feature value or the second feature value corresponding to each pixel in the plurality of sample images to obtain a similarity index of the feature image corresponding to the plurality of sample images respectively.
• Please refer toFIG. 10B, which is a schematic diagram showing afeature image96. The feature image can be processed with each of the sample images for further calculation to obtain a corresponding similarity index C_uv. The calculation is based on normalized correlation matching, as described in equation (1). Normalized correlation matching is aimed at calculating the relation between the feature image and the sample image, wherein the standard deviation of the greyscale value of each image is regarded as a vector and is multiplied with the weight value so as to determine the optimal location. The standard correlation value is within the range between −1 and 1 with higher similarity as it gets closer to 1. When C_uvreaches its maximum, an optimal location is achieved.
• $\begin{array}{cc}{C}_{\mathrm{uv}}=\frac{\sum \left(u_i-\stackrel{\_}{u}\right)\left(v_i-\stackrel{\_}{v}\right)\times w_i}{{\left[\sum {\left(u_i-\stackrel{\_}{u}\right)}^2\sum {\left(v_i-\stackrel{\_}{v}\right)}^2\right]}^{1/2}}& \left(1\right)\end{array}$
• wherein u_iis the greyscale value of each pixel in the sample image, while v_iis the greyscale value of each pixel in the feature image, i.e., the third feature value. Moreover, ū is the average greyscale value of all the pixels in the sample image, whilev is the average greyscale value of all the pixels in the feature image. w_iis the weight value of the pixels in the standard image region and the non-standard image region in the sample image. The weight value of pixels in the other region is 1.
• Based on equation (1), a calculation is performed on each pixel inFIG. 10B and each pixel in the sample image. For example,FIG. 10B and the sample image inFIG. 9C (representing digit 1) and the sample image inFIG. 9D (representing digit 0) are calculated to obtain the similarity index C_uvof the feature image inFIG. 10B corresponding toFIG. 9C andFIG. 9D. Referring toFIG. 8, after obtaining the similarity index, steps43 and44 are performed to acquire the feature image from each character in therecombined image95.Step42 is then repeated to perform identification.Step45 collects a plurality of similarity indexes with respect to the feature image compared with the plurality of sample images. In the present step, the similarity indexes are sorted from the identification result with highest possibility to the identification result with lowest possibility. Finally, instep46, the plurality of similarity indexes are sorted and at least one of comparison results is output.
• Referring toFIG. 3B, since there are 7 characters in the identification mark, the result shown inFIG. 11 can be obtained after the flowchart of themethod4 for identification is performed. InFIG. 11, four possible results are shown. Each result represents one possible combination of characters on the license plate. Each character in the first possible result has the highest similarity, which is followed by the second, the third and the fourth possible results. Taking the first possible result for example, the characters on the license plate are possibly 6095-OA, wherein the first character “6” has a similarity index of 72, the second character “0” has a similarity index of 52, the third character “9” has a similarity index of 67, the fourth character “5” has a similarity index of 72, the fifth character is “-”, sixth character “O” has a similarity index of 63, and the seventh character “A” has a similarity index of 76. Certainly, the user can also determine other combinations of characters on the license plate number according to the results inFIG. 11 and visual estimation on the image to be identified.
• In image identification, images of impossible characters or digits can be excluded according to various combinations that form the identification marks. For example, in one embodiment, the identification mark can be formed as a combination of 4 leading digits and 2 following letters (as shown inFIG. 3A) with a “-” therebetween. In another identification mark, 2 leading letters and 4 follwing digits are combined, with a “-” therebetween. In the present embodiment, there are two kinds of combinations to exemplify the license plates. Therefore, images of impossible characters or digits can be excluded according to the relative locations of the feature images in the identification mark so as to increase identification efficiency.
• Please refer toFIG. 12, which is a schematic diagram of a system for image acquiring and identification according to the present invention. Thesystem6 is capable of implementing the flowchart inFIG. 1,FIG. 5 orFIG. 8 for image identification and identification result output. Thesystem6 comprises adatabase60, animage processing unit61, an identification andoutput unit62, a plurality ofimage acquiring units63 and animage input unit64. Thedatabase60 is capable of providing a plurality of sample images. The plurality ofimage acquiring units63 are electrically connected to theimage processing unit61. Eachimage acquiring unit63 is capable of acquiring an image of an object and transmits the image to theimage processing unit61 for identification. In the present embodiment, each of theimage acquiring units63 is capable of acquiring dynamic or static images of the object. The image provides an identification region for carrier identification. The identification region comprises identification information. The image acquiring units may be CCD or CMOS image acquiring units, but not limited thereto. The object may be a carrier with an identification mark thereon, for example, the license plate number of a car. Moreover, the object may also be a word, a character, a digit or combinations thereof.
• Theimage input unit64 is capable of receiving and transmitting the plurality of images acquired by theimage acquiring unit63 to theimage processing unit61. Theimage processing unit61 comprises afeature acquiring unit610, arecombination unit611, an enhancingunit612 and an identification andcomparison unit613. Thefeature acquiring unit610 is capable of acquiring features in the region of interest in a standard image and acquiring a feature region corresponding to the region of interest according to acquired features in other regions. The standard image is formed by choosing one image from the plurality of images. Therecombination unit611 performs an image recombination process according to the plurality of feature regions and the region of interest to form a recombined image. The recombined image is formed as disclosed inFIG. 1 orFIG. 5. The enhancingunit612 is capable of improving the recombined image to enhance the contrast, brightness or the edge features of the recombined image.
• The identification and comparison unit4111 performs step23 inFIG. 1 to compare the feature image with the sample image to obtain the plurality of similarity indexes corresponding thereto, and sorts the plurality of similarity indexes to output at least one of comparison results. The identification andoutput unit42 is electrically connected to theprocessing unit41 to output the comparison result identified by theprocessing unit41. The output from the identification andoutput unit42 is as shown inFIG. 8A, which is capable of allowing the user to know the identification results displayed on a display.
• The identification andcomparison unit613 is electrically connected to the enhancingunit612 to identify the recombined image. The identification andcomparison unit613 performs a calculation on each pixel in the feature image and each pixel in the plurality of sample images to obtain a similarity index of the feature image corresponding to the plurality of sample images respectively according to the flowchart inFIG. 8, and further collects a plurality of similarity indexes with respect to the feature image compared with the plurality of sample images. Theoutput unit62 is electrically connected to theprocessing unit61 to output at least one of comparison results from theprocessing unit61.
• Accordingly, the present invention discloses a method for image processing and identification and a system for image acquiring and identification with enhanced efficiency and precision. Therefore, the present invention is useful, novel and non-obvious.
• Although this invention has been disclosed and illustrated with reference to particular embodiments, the principles involved are susceptible for use in numerous other embodiments that will be apparent to persons skilled in the art. This invention is, therefore, to be limited only as indicated by the scope of the appended claims.

Claims (24)

1. A method for image recombination of a plurality of images, comprising steps of:

acquiring a plurality of images;

determining a region of interest in an image from the plurality of images and acquiring features in the region of interest;

acquiring from other regions a feature region corresponding to the region of interest according to the acquired features; and

performing an image recombination process to form a recombined image according to a plurality of feature regions and the region of interest.

2. The method for image recombination of a plurality of images as recited inclaim 1, further comprising a step of performing an image enhancing process on the recombined image.

3. The method for image recombination of a plurality of images as recited inclaim 1, wherein the image recombination process comprises steps of:

performing an operation process on corresponding pixels in the plurality of feature regions and the region of interest; and

forming the recombined image after the operation process.

4. The method for image recombination of a plurality of images as recited inclaim 3, wherein the corresponding pixels are averaged in the operation process.

5. The method for image recombination of a plurality of images as recited inclaim 4, further comprising a step of performing a histogram equalization process on the recombined image.

6. The method for image recombination of a plurality of images as recited inclaim 3, wherein the operation process further comprises steps of:

performing an inversing operation on the pixels in the plurality of feature regions and the pixels in the region of interest respectively;

summing the pixels in the plurality of feature regions and the pixels in the region of interest to form the recombined image; and

performing an inversing operation on the recombined image.

7. The method for image recombination of a plurality of images as recited inclaim 1, wherein the region of interest is determined by choosing from the plurality of images an image as a standard image and choosing from the standard image a region as the region of interest.

8. The method for image recombination of a plurality of images as recited inclaim 1, further comprising a step of normalizing the plurality of images.

9. A method for image identification, comprising steps of:

acquiring a plurality of images;

determining a region of interest in an image from the plurality of images and acquiring features in the region of interest;

acquiring from other regions a feature region corresponding to the region of interest according to the acquired features; and

performing an image recombination process to form a recombined image according to a plurality of feature regions and the region of interest; and

identifying the recombined image.

10. The method for image identification as recited inclaim 9, wherein the step of identifying the recombined image further comprises steps of:

acquiring at least one feature image in the recombined image;

providing a database comprising a plurality of sample images therein, each having respectively a standard image region and at least a non-standard image region, wherein the standard image region has pixels corresponding to a first feature value respectively, and the non-standard image region has pixels corresponding to a second feature value respectively;

performing a calculation on a third feature value of each pixel in the feature image and the first feature value or the second feature value corresponding to each pixel in the plurality of sample images to obtain a similarity index of the feature image corresponding to the plurality of sample images respectively;

collecting a plurality of similarity indexes with respect to the feature image compared with the plurality of sample images; and

sorting the plurality of similarity indexes and outputting at least one of comparison results.

11. The method for image identification as recited inclaim 10, wherein the calculation is based on normalized correlation matching.

12. The method for image identification as recited inclaim 10, further comprising a step of normalizing the feature image to adjust the size and the angle of the feature image so that the size of the feature image is identical to the size of the sample image after acquiring the feature image from the image.

13. The method for image identification as recited inclaim 10, wherein each sample image corresponds to an image of a digit or a character.

14. The method for image identification as recited inclaim 10, wherein the first feature value and the second feature value are respectively a combination of a weight value and a greyscale value, and the third feature value is a greyscale value.

15. The method for image identification as recited inclaim 9, wherein the recombined image comprises information regarding an identification mark on a carrier.

16. The method for image identification as recited inclaim 9, wherein the image recombination process comprises steps of:

performing an operation process on corresponding pixels in a plurality of feature regions and the region of interest; and

forming the recombined image after the operation process.

17. The method for image identification as recited inclaim 16, wherein the corresponding pixels are averaged in the operation process.

18. The method for image identification as recited inclaim 17, further comprising a step of performing a histogram equalization process on the recombined image.

19. The method for image identification as recited inclaim 16, wherein the operation process further comprises steps of:

performing an inversing operation on the pixels in the plurality of feature regions and the pixels in the region of interest respectively;

summing the pixels in the plurality of feature regions and the pixels in the region of interest to form the recombined image; and

performing an inversing operation on the recombined image.

20. A system for image acquiring and identification, comprising:

an image input unit capable of providing a plurality of images;

an image processing unit coupled to the image input unit, the image processing unit further comprising:

a feature acquiring unit capable of acquiring features in a region of interest on an image from the plurality of images and acquiring from other regions a feature region corresponding to the region of interest according to the acquired features in other regions;

a recombination unit capable of performing an image recombination process according to the plurality of feature regions and the region of interest to form a recombined image.

21. The system for image acquiring and identification as recited inclaim 20, wherein the image processing unit further comprises:

a database comprising a plurality of sample images therein, each having respectively a standard image region and at least a non-standard image region, wherein the standard image region has pixels corresponding to a first feature value respectively, and the non-standard image region has pixels corresponding to a second feature value respectively; and

an identification and comparison unit coupled to the recombination unit to identify the recombined image, the identification and comparison unit performing a calculation on each pixel in the feature image and each pixel in the plurality of sample images to obtain a similarity index of the feature image corresponding to the plurality of sample images respectively, the identification and comparison unit further collecting a plurality of similarity indexes with respect to the feature image compared with the plurality of sample images.

22. The system for image acquiring and identification as recited inclaim 20, wherein the image processing unit is further connected to an output unit to sort the plurality of similarity indexes and output at least one of comparison results.

23. The system for image acquiring and identification as recited inclaim 20, wherein the recombination unit is further connected to an enhancing unit to enhance the recombined image.

24. The system for image acquiring and identification as recited inclaim 20, wherein the recombination unit is capable of normalizing the plurality of images for an image recombination process.

Images