Color card image segmentation method based on peak detection and DBSCAN clustering

Abstract

The invention provides a color card image segmentation method based on peak detection and DBSCAN clustering, and the method comprises the steps: carrying out the image enhancement of a format color card image, and obtaining an enhanced format color card image; smoothing the enhanced format color card image by adopting a Gaussian blur method to obtain an image feature value after Gaussian blur; performing convolution calculation on the feature value of the image after Gaussian blur along the vertical direction to obtain a column feature value of the image, and performing column segmentation on the image based on the column feature value; performing convolution calculation on the feature value of the image after Gaussian blur along the horizontal direction to obtain a row feature value of the image, and performing row segmentation on the image based on the row feature value; and screening the small color card blocks obtained through column segmentation and row segmentation by a density-based noisy spatial clustering algorithm to obtain all correct color card blocks. The color card image segmentation method has wide applicability and flexibility, and can meet color card image segmentation requirements in different fields and scenes.

Description

Technical Field

[0001] The present invention relates to an image processing method, and in particular to a segmentation method of a color card image based on peak detection and DBSCAN clustering. Background Art

[0002] In today's manufacturing globalization, informatization and digitalization environment, the leather, fur, feather and their products and shoemaking industries, as important parts of traditional industries, are facing huge demands for digital transformation. This huge market size shows the industry's huge demand for innovative technologies.

[0003] In this context, with the rapid development of computer technology, automatic image processing technology based on computer vision has brought new opportunities for the digitization of color cards. The digitization of physical color card images has brought far-reaching impacts on the management, design, production and sustainable development of color card materials. The digitization process of color cards enables each piece of color card material to be accurately recorded, classified and tracked, greatly improving the accuracy and efficiency of color card material inventory management, and providing reliable data support for production planning and material procurement; based on the digitization of color cards, it can also quickly generate a variety of design solutions for various application products by screening and combining a large number of color card material samples, reducing the cost problems caused by repeated modifications due to mismatched color card materials, and greatly improving the design efficiency and innovation capabilities of various products in the industry; color card digitization can also help users save time and reduce costs in the process of finding materials, selecting materials and comparing through accurate color card image retrieval technology. Therefore, how to segment the current paper version of the color card style and convert it into several color card block images is the key link in the digitization of color cards. Summary of the invention

[0004] In view of the above situation, the main purpose of the present invention is to propose a segmentation method for color card images based on peak detection and DBSCAN clustering to solve the above technical problems.

[0005] The present invention proposes a segmentation method for color card images based on peak detection and DBSCAN clustering, and the method comprises the following steps: Step 1, performing image enhancement on the layout color card image to obtain an enhanced layout color card image; Step 2: The enhanced layout color card image is smoothed by using a Gaussian blur method to obtain the image feature value after Gaussian blur; Step 3: along the vertical direction, a sliding window is used to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the column feature values ​​of the image; Obtaining the maximum value of each local column feature and two adjacent local minimum values ​​of each local column feature based on the column feature value of the image; Based on the maximum value of each local column feature and two adjacent local minimum values ​​of each local column feature, the boundary of the layout color card image column is identified by saliency discrimination, and the layout color card image is segmented by column; Step 4: In the horizontal direction, a sliding window is used to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the row feature values ​​of the image; Obtaining the maximum value of each local row feature and two adjacent local minimum values ​​of each local row feature based on the row feature value of the image; Based on the maximum value of each local row feature and two adjacent local minimum values ​​of each local row feature, the boundary of the layout color card image row is identified by saliency discrimination, and the layout color card image is segmented by row; Step 5: The color card blocks obtained by column segmentation and row segmentation are screened by a density-based spatial clustering algorithm with noise to obtain all correct color card blocks.

[0006] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention realizes efficient and accurate segmentation of paper format color card images by combining Gaussian fuzzy preprocessing, eigenvalue peak detection and DBSCAN clustering technology. Compared with traditional manual cutting, the present invention has a high accuracy rate and greatly improves efficiency; 2. The image processing method of the present invention is not only applicable to standard paper format color cards, but also can process color card images with different sizes, shapes and arrangements. Through size normalization processing, the system can automatically adjust the size of the input image to adapt to different processing requirements. The image processing method of the present invention has wide applicability and flexibility, and can meet the color card image segmentation requirements in different fields and scenarios.

[0007] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description or learned through embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a flow chart of the segmentation method of the color card image based on peak detection and DBSCAN clustering proposed by the present invention. DETAILED DESCRIPTION

[0009] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0010] These and other aspects of the embodiments of the present invention will be apparent with reference to the following description and accompanying drawings. In these descriptions and accompanying drawings, some specific implementations of the embodiments of the present invention are specifically disclosed to represent some ways of implementing the principles of the embodiments of the present invention, but it should be understood that the scope of the embodiments of the present invention is not limited thereto.

[0011] See also Figure 1 The embodiment of the present invention proposes a segmentation method for a color card image based on peak detection and DBSCAN clustering, the method comprising the following steps: Step 1, performing image enhancement on the layout color card image to obtain an enhanced layout color card image; In step 1, the layout color card image is enhanced to obtain an enhanced layout color card image. The specific steps are as follows: Convert the paper layout color card style into a layout color card image through an image acquisition device; Perform resolution normalization on the collected layout color card images; Furthermore, the constructed layout color card image has a 24-bit color depth, including but not limited to the image adopting the JPG format, including but not limited to the resolution of 3500*5000 pixels. The present invention automatically converts the collected layout color card image into an image with a resolution of 3000*4500 pixels. This standard ensures that the layout color card image can contain sufficient feature details.

[0012] Step 2: The enhanced layout color card image is smoothed by using a Gaussian blur method to obtain the image feature value after Gaussian blur; In step 2, the enhanced layout color card image is smoothed by using the Gaussian blur method to obtain the image feature value after Gaussian blur, wherein the weight of the Gaussian function is calculated using the position of the pixel points in the enhanced layout color card image, and the corresponding process has the following relationship: ; in, represents the weight of the Gaussian function, It means that after the natural exponential function calculation, represents pi, represents the standard deviation of the normal distribution, Indicates the horizontal position of the pixel. Indicates the vertical position of the pixel; The enhanced layout color card image is smoothed by Gaussian blur method to obtain the image feature value after Gaussian blur. The image feature value after Gaussian blur is calculated by using the weight of Gaussian function. The corresponding process has the following relationship: ; in, represents the image feature value after Gaussian blur, represents the unprocessed image feature value, Represents the three channels of a color image, represents the radius of the Gaussian kernel, Indicates that the Gaussian kernel is located in the layout color card image The weight of Indicates the horizontal offset. Indicates the vertical offset.

[0013] Furthermore, the present invention performs Gaussian blurring and smoothing processing on the enhanced layout color card image. Since the color, pattern, size, texture and other information of the color card are complex, in order to reduce its influence on boundary recognition, the present invention uses Gaussian blurring to reduce noise.

[0014] Step 3: along the vertical direction, a sliding window is used to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the column feature values ​​of the image; Obtaining the maximum value of each local column feature and two adjacent local minimum values ​​of each local column feature based on the column feature value of the image; Based on the maximum value of each local column feature and two adjacent local minimum values ​​of each local column feature, the boundary of the layout color card image column is identified through significance judgment, and the layout color card image is segmented into columns.

[0015] In step 3, a sliding window is used in the vertical direction to perform convolution calculation on the image eigenvalues ​​after Gaussian blur in the window to obtain the column eigenvalues ​​of the image. The corresponding relationship in the process is: ; in, represents the column feature value of the image, Represents the width of the pooled layout color card image, Represents the length of the pooled layout color card image, Represents the weight of the sliding window; Further, using a 1*3 matrix: Convolution calculation is performed with the enhanced layout color card image: First, a sliding window operation is performed on the three channels (red channel, green channel and blue channel) of the enhanced layout color card image. The operation slides step by step in the vertical direction (from top to bottom) and the horizontal direction (from left to right) with a step size of 1, and the pixel value of each channel in the window is calculated according to the above formula.

[0016] Secondly, the absolute values ​​of the pixel values ​​at the corresponding positions of the three channels are accumulated to obtain a two-dimensional array of size W*H, and the array elements are the accumulated eigenvalues ​​of the corresponding positions.

[0017] Third, the obtained two-dimensional array is accumulated by column feature values ​​to generate a one-dimensional array of layout color card image column features with a length of W. Finally, the local maximum value is calculated on the one-dimensional image column feature array, denoted as , where L_peak is the column where the local maximum value is located, and the two local minima adjacent to the local maximum value are denoted as Lmax , the prominence of the eigenvalue: .

[0018] If the significance is greater than the preset value, the column is the segmentation boundary of the layout color card image column.

[0019] Step 4: In the horizontal direction, a sliding window is used to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the row feature values ​​of the image; Obtaining the maximum value of each local row feature and two adjacent local minimum values ​​of each local row feature based on the row feature value of the image; Based on the maximum value of each local row feature and two adjacent local minimum values ​​of each local row feature, the boundary of the layout color card image row is identified by saliency discrimination, and the layout color card image is segmented by row; In step 4, a sliding window is used in the horizontal direction to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the row feature values ​​of the image. The relationship between the corresponding process is: ; in, Indicates the width of the layout color card column image after column division. Represents the row features of the image, Represents the weight of the sliding window; Further, using a 2*1 matrix: Calculate with the original image, select the step size of the sliding window as 1, and fill it with 0.

[0020] First, a sliding window operation is performed on the three channels (red channel, green channel and blue channel) of the enhanced layout color card image. The operation slides step by step in the horizontal direction (from left to right) and vertical direction (from top to bottom) with a step size of 1, and the pixel value of each channel in the window is calculated according to the above formula.

[0021] Secondly, the absolute values ​​of the pixel values ​​at the corresponding positions of the three channels are accumulated to obtain a two-dimensional array of size VW*H, and the array elements are the accumulated eigenvalues ​​of the corresponding positions.

[0022] Third, the obtained two-dimensional array is accumulated by row feature values ​​to generate a one-dimensional array of row features of the color card column image with a length of H. Finally, the local maximum value is calculated on the one-dimensional image column feature array, denoted as , where L_peak is the row where the local maximum is located, and the two local minima adjacent to the local maximum are denoted as Lmix , the prominence of the eigenvalue: .

[0023] If the significance is greater than the preset value, the column is the dividing boundary of the layout color card column image row.

[0024] Step 5: Screen the color card blocks obtained by column segmentation and row segmentation using a density-based spatial clustering algorithm with noise to obtain all correct color card blocks; In step 5, the color card blocks obtained by column segmentation and row segmentation are screened by a density-based spatial clustering algorithm with noise to obtain all correct color card blocks. The density-based spatial clustering algorithm with noise includes two parameters: domain radius and minimum number of points. The definition of the domain radius is: ; in, Represents the set of all color card blocks, and All represent small pieces of color cards. Indicates the absolute value of the area difference between the two color card blocks. Indicates the radius of the field, Indicates is the set of points in the area of ​​the cluster center, express The area of express area; The definition of the minimum number of points is: ; in, Indicates the minimum number of points, Indicates is the number of points contained in the area of ​​the cluster center; Furthermore, in the present invention, The value is 1000. The value is 10, which means that as long as there are 10 other color card blocks in the neighborhood and the pixel area difference is less than 1000, it can be processed as a core point.

[0025] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0026] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0027] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A segmentation method for color card images based on peak detection and DBSCAN clustering, characterized in that: The method comprises the following steps: Step 1, performing image enhancement on the layout color card image to obtain an enhanced layout color card image; Step 2: The enhanced layout color card image is smoothed by using a Gaussian blur method to obtain the image feature value after Gaussian blur; Step 3: along the vertical direction, a sliding window is used to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the column feature values ​​of the image; Obtaining the maximum value of each local column feature and two adjacent local minimum values ​​of each local column feature based on the column feature value of the image; Based on the maximum value of each local column feature and two adjacent local minimum values ​​of each local column feature, the boundary of the layout color card image column is identified by saliency discrimination, and the layout color card image is segmented by column; Step 4: In the horizontal direction, a sliding window is used to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the row feature values ​​of the image; Obtaining the maximum value of each local row feature and two adjacent local minimum values ​​of each local row feature based on the row feature value of the image; Based on the maximum value of each local row feature and two adjacent local minimum values ​​of each local row feature, the boundary of the layout color card image row is identified by saliency discrimination, and the layout color card image is segmented by row; Step 5: The color card blocks obtained by column segmentation and row segmentation are screened by a density-based spatial clustering algorithm with noise to obtain all correct color card blocks.

2. The segmentation method of color card image based on peak detection and DBSCAN clustering according to claim 1, characterized in that: In step 1, the layout color card image is enhanced to obtain an enhanced layout color card image. The specific steps are as follows: Convert the paper layout color card style into a layout color card image through an image acquisition device; Perform resolution normalization on the collected layout color card images.

3. The segmentation method of color card image based on peak detection and DBSCAN clustering according to claim 2 is characterized in that: In step 2, the enhanced layout color card image is smoothed by using a Gaussian blur method to obtain the image feature value after Gaussian blur, wherein the weight of the Gaussian function is calculated using the position of the pixel points in the enhanced layout color card image, and the corresponding process has the following relationship: ； in, represents the weight of the Gaussian function, It means that after the natural exponential function calculation, represents pi, represents the standard deviation of the normal distribution, Indicates the horizontal position of the pixel. Indicates the vertical position of the pixel.

4. The segmentation method of color card image based on peak detection and DBSCAN clustering according to claim 3 is characterized in that: In step 2, the enhanced layout color card image is smoothed by using a Gaussian blur method to obtain the image feature value after Gaussian blur, wherein the image feature value after Gaussian blur is calculated using the weight of the Gaussian function, and the corresponding process has the following relationship: ； in, represents the image feature value after Gaussian blur, represents the unprocessed image feature value, Represents the three channels of a color image, represents the radius of the Gaussian kernel, Indicates that the Gaussian kernel is located in the layout color card image The weight of Indicates the horizontal offset. Indicates the vertical offset.

5. The segmentation method of color card image based on peak detection and DBSCAN clustering according to claim 4 is characterized in that: In step 3, a sliding window is used in the vertical direction to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the column feature values ​​of the image. The corresponding process has the following relationship: ； in, represents the column feature value of the image, Indicates the width of the pooled layout color card image, Represents the length of the pooled layout color card image, Represents the weight of the sliding window.

6. The segmentation method of color card image based on peak detection and DBSCAN clustering according to claim 5 is characterized in that: In step 4, a sliding window is used in the horizontal direction to perform convolution calculation on the image feature values ​​after Gaussian blur in the window to obtain the row feature values ​​of the image. The relationship between the corresponding process is: ； in, Indicates the width of the layout color card column image after column division. Represents the row features of the image, Represents the weight of the sliding window.

7. The method for segmenting color card images based on peak detection and DBSCAN clustering according to claim 6, characterized in that: In step 5, the color card blocks obtained by column segmentation and row segmentation are screened by a density-based spatial clustering algorithm with noise to obtain all correct color card blocks, wherein the density-based spatial clustering algorithm with noise includes two parameters: domain radius and minimum number of points. The definition of the domain radius is: ； in, Represents the set of all color card blocks, and All represent small pieces of color cards. Indicates the absolute value of the area difference between the two color card blocks. Indicates the radius of the field, Indicates is the set of points in the area of ​​the cluster center, express The area of express area.

8. The method for segmenting color card images based on peak detection and DBSCAN clustering according to claim 7, characterized in that: In step 5, the color card blocks obtained by column segmentation and row segmentation are screened by a density-based spatial clustering algorithm with noise to obtain all correct color card blocks; wherein the density-based spatial clustering algorithm with noise includes two parameters: domain radius and minimum number of points. The definition of the minimum number of points is: ； in, Indicates the minimum number of points, Indicates is the number of points contained in the area of ​​the cluster center.

Images