CN112330611B - FPGA-oriented high-parallelism light spot segmentation method - Google Patents

Abstract

The invention relates to a high-parallelism facula segmentation method facing FPGA, which belongs to the field of image processing, and sequentially reads in 2X 2 pixel areas from left to right and from top to bottom, the 2X 2 pixel areas are marked as a tuple, a current tuple is marked as c, an upper tuple is marked as t, a right upper tuple is marked as tr, a left Fang Yuan group is marked as l, each tuple comprises 1, 2, 3 and 4 pixels, and based on the characteristic of parallelism of FPGA, the method comprises the following main steps: the spot numbers of the pixels of each tuple are determined according to the connected areas, and data are integrated according to the number relevance; the invention has the characteristics of flexibility, parallelism and integration, can realize image acquisition and processing, and greatly improves the speed of image processing.

Description

A high parallel spot segmentation method for FPGA

Technical Field

The invention relates to the field of image processing, and in particular to a high-parallelism spot segmentation method oriented to FPGA.

Background technique

Light spot is a common image information in image processing. The centroid of the light spot is an important feature of the light spot image and is often used to locate target points in the field of visual measurement. The premise of high-precision positioning of the centroid of the light spot is to accurately segment the light spot.

There are two types of spot segmentation methods: software-based methods and hardware-based methods. Most software-based methods segment the entire image after image acquisition is completed. However, when the image resolution is high or there are many spots in the image, the real-time performance of this type of method is difficult to guarantee. In addition, this type of method generally requires a large data storage space to save the image to be processed, especially in embedded systems where the storage space itself is limited. Existing hardware-based methods can realize the acquisition and processing of low-resolution images, and the required storage space is small, but this type of method only processes one pixel at a time. Therefore, when the image resolution is high, the efficiency of this type of method is also difficult to guarantee.

Summary of the invention

In view of the problems existing in the prior art, the present invention discloses a high-parallelism light spot segmentation method for FPGA. The method processes 4 pixels simultaneously each time and regards the current 4 pixels as a region. By analyzing the properties of the regions above, to the upper right and to the left of the region, the light spot properties of all pixels in the current region are judged, and finally all the light spots in the image are segmented. The method solves the defects in the prior art.

The present invention is achieved in that:

A high-parallelism spot segmentation method for FPGA is characterized in that a 2×2 pixel area is read in sequence from left to right and from top to bottom, the 2×2 pixel area is recorded as a tuple, the current tuple is recorded as c, the upper tuple is recorded as t, the upper right tuple is recorded as tr, the left tuple is recorded as l, each tuple contains 4 pixels, 1, 2, 3, 4, and 4 pixels in the current tuple are processed in parallel each time, the spot attributes of all pixels in the current tuple are judged by analyzing the attributes of the tuples above, upper right and left of the current tuple, and finally all the spots in the image are segmented; the specific steps include: determining the spot number of each tuple pixel according to the connected area, integrating data according to the number correlation, and after all tuples in the image are processed, the pixels with the same pixel number belong to the same spot.

Furthermore, the specific steps of determining the spot number of each tuple pixel are:

Step 1: Determine whether the c1 pixel in the current tuple is a light spot pixel: if the grayscale of the c1 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 2: If the c1 pixel in the current tuple is a spot pixel, determine the spot number label_c1 of the c1 pixel according to the connected area; determine the spot number label_c1 of the c1 pixel according to the pixel numbers above, above right, above right, to the left, and below left of the c1 pixel (label_t3, label_t4, label_tr3, label_l2, label_l4);

Step 3: Determine whether the c2 pixel in the current tuple is a light spot pixel: if the grayscale of the c2 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 4: If the c2 pixel in the current tuple is a spot pixel, determine the spot number label_c2 of the c2 pixel according to the connected area. Determine the spot number label_c2 of the c2 pixel according to the pixel numbers above, above right, to the left, and below the left left of the c2 pixel (label_t4, label_tr3, label_c1, label_l4);

Step 5: Determine whether the c3 pixel in the current tuple is a light spot pixel: if the grayscale of the c3 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 6: If the c3 pixel in the current tuple is a spot pixel, determine the spot number label_c3 of the c3 pixel according to the connected area; determine the spot number label_c3 of the c3 pixel according to the pixel numbers above, above right, and to the left of the c3 pixel (label_c1, label_c2, label_l4);

Step 7: Determine whether the c4 pixel in the current tuple is a light spot pixel: if the grayscale of the c4 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 8: If the c4 pixel in the current tuple is a spot pixel, determine the spot number label_c4 of the c4 pixel according to the connected area; determine the spot number label_c4 of the c4 pixel according to the pixel numbers above and to the left of the c4 pixel (label_c2, label_c3);

Further, in the steps 1 to 8, steps 1 and 2, steps 3 and 4, steps 5 and 6, and steps 7 and 8 are performed in parallel, wherein:

The specific steps of determining the spot number label_c1 of the c1 pixel are as follows: if the upper pixel is a spot pixel, label_c1=label_t3; otherwise, if the upper right pixel is a spot pixel, label_c1=label_t4; otherwise, if the grayscales of the c2 pixel and the tr3 pixel are both greater than the threshold T, label_c1=label_tr3; otherwise, if the left pixel is a spot pixel, label_c1=label_l2; otherwise, if the lower left pixel is a spot pixel, label_c1=label_l4; otherwise, the c1 pixel is a new spot pixel and is numbered with a new label;

The specific steps of determining the spot number label_c2 of the c2 pixel are: if the upper pixel is a spot pixel, label_c2=label_t4; otherwise, if the upper right pixel is a spot pixel, label_c2=label_tr3; otherwise, if the left pixel is a spot pixel, label_c2=label_c1. In this process, it is necessary to determine whether the c1 pixel is a spot pixel through steps 1 and 2, and determine the number of the c1 pixel; otherwise, if the grayscales of the c3 pixel and the l4 pixel are both greater than the threshold T, label_c2=label_l4; otherwise, the c2 pixel is a new spot pixel and is numbered with a new label;

The specific steps of determining the spot number label_c3 of the c3 pixel are as follows: if the upper pixel is a spot pixel, label_c3=label_c1. In this process, it is necessary to determine whether the c1 pixel is a spot pixel through steps 1 and 2, and determine the number of the c1 pixel; otherwise, if the upper right pixel is a spot pixel, label_c3=label_c2. In this process, it is necessary to determine whether the c2 pixel is a spot pixel through steps 3 and 4, and determine the number of the c2 pixel; otherwise, if the left pixel is a spot pixel, label_c3=label_l4; otherwise, the c3 pixel is a new spot pixel, and a new label is used to number it;

The specific steps for determining the spot number label_c4 of the c4 pixel are: if the upper pixel is a spot pixel, then label_c4=label_c2. In this process, it is necessary to determine whether the c2 pixel is a spot pixel through steps 3 and 4, and determine the number of the c2 pixel; otherwise, if the left pixel is a spot pixel, then label_c4=label_c3. In this process, it is necessary to determine whether the c3 pixel is a spot pixel through steps 5 and 6, and determine the number of the c3 pixel; otherwise, the c4 pixel is a new spot pixel and is numbered with a new label.

Furthermore, the process of integrating data according to the association of numbers is as follows: after all pixels in the current tuple are numbered, 3 consecutive tuples are taken from the tuple to the left, and the numbers in the 3 tuples are compared. If they are different, the larger number is changed to the smallest number in the 3 tuples.

Compared with the prior art, the present invention has the following beneficial effects: the method of the present invention can realize simultaneous acquisition and processing of high-resolution images, and the method utilizes the high parallelism of FPGA to process four pixels in the same area in parallel each time, which greatly improves the image processing speed. At the same time, based on the online processing characteristics of the method, the use of a relatively small memory can meet the demand.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of light spot tuple division in a high-parallelism light spot segmentation method for FPGA of the present invention;

FIG2 is a flowchart of determining the c1 pixel number in a high-parallel spot segmentation method for FPGA of the present invention;

FIG3 is a flow chart of determining the c2 pixel number in a high-parallelism spot segmentation method for FPGA of the present invention;

FIG4 is a flow chart of determining the c3 pixel number in a high-parallelism spot segmentation method for FPGA of the present invention;

FIG5 is a flow chart of determining the c4 pixel number in a high-parallelism spot segmentation method for FPGA of the present invention;

FIG6 is a schematic diagram of data numbering before data integration according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of numbering after data integration in an embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail by enumerating examples below. It should be noted that the specific implementation described here is only used to explain the present invention and is not used to limit the present invention.

The present invention provides a high-parallel spot segmentation method for FPGA: read in a 2×2 pixel area from left to right and from top to bottom, record the 2×2 pixel area as a tuple, record the current tuple as c, the upper tuple as t, the upper right tuple as tr, the left tuple as l, and each tuple contains 4 pixels, 1, 2, 3, and 4, as shown in Figure 1. Based on the characteristics of FPGA parallelism, the method processes 4 pixels in the current tuple in parallel each time, and mainly includes the following steps: determine the spot number of each tuple pixel according to the connected area, and integrate data according to the number correlation.

The steps to determine the spot number of each tuple pixel are as follows:

Step 1: Determine whether the c1 pixel in the current tuple is a light spot pixel. If the grayscale of the c1 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 2: If the c1 pixel in the current tuple is a spot pixel, the spot number label_c1 of the c1 pixel is determined according to the connected area. The spot number label_c1 of the c1 pixel is determined according to the pixel numbers above, above right, above right, to the left, and below left of the c1 pixel (label_t3, label_t4, label_tr3, label_l2, label_l4). Specifically: if the upper pixel is a spot pixel, label_c1 = label_t3; otherwise, if the upper right pixel is a spot pixel, label_c1 = label_t4; otherwise, if the grayscale of the c2 pixel and the tr3 pixel is greater than the threshold T, label_c1 = label_tr3; otherwise, if the left pixel is a spot pixel, label_c1 = label_l2; otherwise, if the lower left pixel is a spot pixel, label_c1 = label_l4; otherwise, the c1 pixel is a new spot pixel and is numbered with a new label. The process flow chart is shown in Figure 2.

Step 3: Determine whether the c2 pixel in the current tuple is a light spot pixel. If the grayscale of the c2 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 4: If the c2 pixel in the current tuple is a spot pixel, the spot number label_c2 of the c2 pixel is determined according to the connected area. The spot number label_c2 of the c2 pixel is determined according to the pixel numbers above, above right, to the left, and below the left left of the c2 pixel (label_t4, label_tr3, label_c1, label_l4). Specifically: if the upper pixel is a spot pixel, label_c2 = label_t4; otherwise, if the upper right pixel is a spot pixel, label_c2 = label_tr3; otherwise, if the left pixel is a spot pixel, label_c2 = label_c1. In this process, it is necessary to determine whether the c1 pixel is a spot pixel through steps 1 and 2, and determine the number of the c1 pixel; otherwise, if the grayscale of the c3 pixel and the l4 pixel is greater than the threshold T, label_c2 = label_l4; otherwise, the c2 pixel is a new spot pixel and is numbered with a new label. The process flow chart is shown in Figure 3.

Step 5: Determine whether the c3 pixel in the current tuple is a light spot pixel. If the grayscale of the c3 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 6: If the c3 pixel in the current tuple is a spot pixel, the spot number label_c3 of the c3 pixel is determined according to the connected area. The spot number label_c3 of the c3 pixel is determined according to the pixel numbers above, above right, and to the left of the c3 pixel (label_c1, label_c2, label_l4). Specifically: if the upper pixel is a spot pixel, label_c3 = label_c1. In this process, it is necessary to determine whether the c1 pixel is a spot pixel through steps 1 and 2, and determine the number of the c1 pixel; otherwise, if the upper right pixel is a spot pixel, label_c3 = label_c2. In this process, it is necessary to determine whether the c2 pixel is a spot pixel through steps 3 and 4, and determine the number of the c2 pixel; otherwise, if the left pixel is a spot pixel, label_c3 = label_l4; otherwise, the c3 pixel is a new spot pixel, and it is numbered with a new label. The process flow chart is shown in Figure 4.

Step 7: Determine whether the c4 pixel in the current tuple is a light spot pixel. If the grayscale of the c4 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0;

Step 8: If the c4 pixel in the current tuple is a spot pixel, the spot number label_c4 of the c4 pixel is determined according to the connected area. The spot number label_c4 of the c4 pixel is determined according to the pixel numbers above and to the left of the c4 pixel (label_c2, label_c3). Specifically: if the upper pixel is a spot pixel, label_c4 = label_c2. In this process, it is necessary to determine whether the c2 pixel is a spot pixel through steps 3 and 4, and determine the number of the c2 pixel; otherwise, if the left pixel is a spot pixel, label_c4 = label_c3. In this process, it is necessary to determine whether the c3 pixel is a spot pixel through steps 5 and 6, and determine the number of the c3 pixel; otherwise, the c4 pixel is a new spot pixel, and it is numbered with a new label. The process flow chart is shown in Figure 5. The process of integrating data according to the number association is as follows: as shown in Figure 6, after all pixels in the current tuple are numbered, 3 consecutive tuples are taken from the tuple to the left, and the numbers in the 3 tuples are compared. If they are different, the larger number is changed to the smallest number in the 3 tuples, as shown in Figure 7. After all tuples in the image are processed, the pixels with the same number belong to the same light spot.

The above description is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be regarded as within the protection scope of the present invention.

Claims (3)

1. A high-parallelism spot segmentation method for FPGA, characterized in that a 2×2 pixel area is read in sequence from left to right and from top to bottom, the 2×2 pixel area is recorded as a tuple, the current tuple is recorded as c, the upper tuple is recorded as t, the upper right tuple is recorded as tr, and the left tuple is recorded as l, each tuple contains 4 pixels, 1, 2, 3, and 4, and 4 pixels in the current tuple are processed in parallel each time, and the spot attributes of all pixels in the current tuple are judged by analyzing the attributes of the tuples above, upper right, and left of the current tuple, and finally all the spots in the image are segmented; the specific steps include: determining the spot number of the pixels in each tuple according to the connected area, integrating data according to the number correlation, and after all tuples in the image are processed, the pixels with the same pixel number belong to the same spot; the specific steps of determining the spot number of the pixels in each tuple are: Step 1: Determine whether the c1 pixel in the current tuple is a light spot pixel: if the grayscale of the c1 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0; Step 2: If the c1 pixel in the current tuple is a spot pixel, determine the spot number label_c1 of the c1 pixel according to the connected area; determine the spot number label_c1 of the c1 pixel according to the pixel numbers above, above right, above right, to the left, and below left of the c1 pixel (label_t3, label_t4, label_tr3, label_l2, label_l4); Step 3: Determine whether the c2 pixel in the current tuple is a light spot pixel: if the grayscale of the c2 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0; Step 4: If the c2 pixel in the current tuple is a spot pixel, determine the spot number label_c2 of the c2 pixel according to the connected area, and determine the spot number label_c2 of the c2 pixel according to the pixel numbers above, above right, to the left, and below the left left of the c2 pixel (label_t4, label_tr3, label_c1, label_l4); Step 5: Determine whether the c3 pixel in the current tuple is a light spot pixel: if the grayscale of the c3 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0; Step 6: If the c3 pixel in the current tuple is a spot pixel, determine the spot number label_c3 of the c3 pixel according to the connected area; determine the spot number label_c3 of the c3 pixel according to the pixel numbers above, above right, and to the left of the c3 pixel (label_c1, label_c2, label_l4); Step 7: Determine whether the c4 pixel in the current tuple is a light spot pixel: if the grayscale of the c4 pixel is greater than the preset threshold T, it is a light spot pixel, otherwise it is a background pixel and is numbered 0; Step 8: If the c4 pixel in the current tuple is a spot pixel, determine the spot number label_c4 of the c4 pixel according to the connected area; determine the spot number label_c4 of the c4 pixel according to the pixel numbers above and to the left of the c4 pixel (label_c2, label_c3); By taking advantage of the high parallelism of FPGA, 4 pixels in the same area are processed in parallel each time, which greatly improves the image processing speed. 2. The high-parallel spot segmentation method for FPGA according to claim 1, characterized in that, in the steps 1 to 8, steps 1 and 2, steps 3 and 4, steps 5 and 6, and steps 7 and 8 are performed in parallel, wherein: The specific steps of determining the spot number label_c1 of the c1 pixel are as follows: if the upper pixel is a spot pixel, label_c1=label_t3; otherwise, if the upper right pixel is a spot pixel, label_c1=label_t4; otherwise, if the grayscales of the c2 pixel and the tr3 pixel are both greater than the threshold T, label_c1=label_tr3; otherwise, if the left pixel is a spot pixel, label_c1=label_l2; otherwise, if the lower left pixel is a spot pixel, label_c1=label_l4; otherwise, the c1 pixel is a new spot pixel and is numbered with a new label; The specific steps of determining the spot number label_c2 of the c2 pixel are: if the upper pixel is a spot pixel, label_c2=label_t4; otherwise, if the upper right pixel is a spot pixel, label_c2=label_tr3; otherwise, if the left pixel is a spot pixel, label_c2=label_c1. In this process, it is necessary to determine whether the c1 pixel is a spot pixel through steps 1 and 2, and determine the number of the c1 pixel; otherwise, if the grayscales of the c3 pixel and the l4 pixel are both greater than the threshold T, label_c2=label_l4; otherwise, the c2 pixel is a new spot pixel and is numbered with a new label; The specific steps of determining the spot number label_c3 of the c3 pixel are as follows: if the upper pixel is a spot pixel, label_c3=label_c1. In this process, it is necessary to determine whether the c1 pixel is a spot pixel through steps 1 and 2, and determine the number of the c1 pixel; otherwise, if the upper right pixel is a spot pixel, label_c3=label_c2. In this process, it is necessary to determine whether the c2 pixel is a spot pixel through steps 3 and 4, and determine the number of the c2 pixel; otherwise, if the left pixel is a spot pixel, label_c3=label_l4; otherwise, the c3 pixel is a new spot pixel, and a new label is used to number it; The specific steps for determining the spot number label_c4 of the c4 pixel are: if the upper pixel is a spot pixel, then label_c4=label_c2. In this process, it is necessary to determine whether the c2 pixel is a spot pixel through steps 3 and 4, and determine the number of the c2 pixel; otherwise, if the left pixel is a spot pixel, then label_c4=label_c3. In this process, it is necessary to determine whether the c3 pixel is a spot pixel through steps 5 and 6, and determine the number of the c3 pixel; otherwise, the c4 pixel is a new spot pixel and is numbered with a new label. 3. According to the high-parallelism spot segmentation method for FPGA described in claim 1, it is characterized in that the process of integrating data according to the number correlation is as follows: after all pixels in the current tuple are numbered, 3 consecutive tuples are taken from the tuple to the left, and the numbers in the 3 tuples are compared. If they are different, the larger number is changed to the smallest number in the 3 tuples.