Image background removal method fusing color and local ternary similar mode characteristics

Abstract

The invention discloses an image background removal method fusing color and local ternary similar mode characteristics. The method comprises steps that S1, a designated target image is acquired, color characteristics of each pixel of the acquired image are respectively calculated, after ternary division of each pixel and an adjacent pixel of the present image is carried out based on LTSP characteristics, intra-LTSP characteristics corresponding to each pixel are acquired through calculation, background modeling is carried out for the target image based on the color characteristics and the intra-LTSP characteristics, and a background model is acquired through establishment; and S2, the established background model is utilized to carry out foreground segmentation of the target image to remove the background image, and model update of the background model is carried out according to the color characteristics of each pixel and the intra-LTSP characteristics. The method is advantaged in that the method is simple for realization, cost is low, adaptability to illumination change of the background is strong, background removal efficiency is high, the effect is good, and the detection missing rate and the false detection rate are low.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to an image background removal method that combines color and local ternary similarity mode features. Background technique [0002] The basic idea of ​​image background removal is to establish a background model, detect moving objects by comparing the difference between the current frame and the background model, and continuously update the background model. The key to background subtraction lies in the establishment and update of the background model, which requires that the background model should reflect the real background as much as possible, and at the same time be able to adapt to the changes of the background scene. [0003] There are currently many modeling methods and implementation algorithms for image background removal, such as the use of mixed Gaussian background model (GMM), non-parametric background model and the use of sample consistency (SACON) algorithm, ...

AI Technical Summary

Problems solved by technology

[0003] There are currently many modeling methods and implementation algorithms for image background removal, such as the use of mixed Gaussian background model (GMM), non-parametric background model and the use of sample consistency (SACON) algorithm, visual background extraction (ViBe) algorithm, etc. , but there are certain defects in each background model and implementation algorithm: GMM can't accurately model the complex dynamic background with only a few Gaussian kernel functions, and the size of the learning rate faces a dilemma, that is, too fast learning rate It will integrate slower-moving targets into the background, and too slow a learning rate is too sensitive to sudden changes in lighting and scenes; a non-parametric background model can avoid the problem of parametric modeling, but for high and low frequency changes at the same time The situation cannot be handled well; based on the sample consistency (SACON) algorithm, by directly taking N frames of video sequences for background modeling, using a first-in-first-out update strategy, but the algorithm must add blob-level processing to eliminate the foreground The "hole" of the target; in background removal, ViBe regards background removal as a classification method. When selecting in the color space, it compares the value of the pixel at the current position with the value of its neighboring pixels. To judge whether the pixel is a foreground pixel or a background pixel, the interference of extreme pixel values ​​can be avoided. The ViBe algorithm can build a background model with only one frame of image. At the same time, it uses a global threshold without the concept of time domain. The model is simple and real-time strong, but there are disadvantages such as long ghost removal time, incomplete detection of static targets, shadowed foreground and moving targets, and many false detections in dynamic scenes. Wrong foreground is generated, and the initial frame is often ghosted

[0005] (1) Use the pixel flickering level to describe the background dynamics, but introduces block-level and frame-level processing when suppressing the dynamic background, and when the moving target appears in the high dynamic background area, it will cause a large missed detection;

[0006] (2) According to the characteristics of biological vision, on the basis of the ViBe algorithm, the mean shift filter is used to blur local details, which has a certain effect on illumination changes and small background disturbances, but has a large amount of calculation and poor real-time performance;

[0007] (3) Time domain information is introduced into the framework of the ViBe algorithm, modeled with continuous N-frame images, and the adaptive threshold is obtained by calculating the statistical information of the model samples. The response time of the disturbed background is slow, and it also gives up the advantage of the single-frame modeling of the ViBe algorithm, which increases the computational complexity and memory overhead;

[0008] (4) A background subtraction algorithm based on ternary pattern features. This model has good adaptability to illumination changes, but the detection target usually has holes caused by a large number of missed detections.

[0009] Chinese patent application CN106127151A discloses a finger vein recognition method and device based on an improved local binary model. By constructing an improved local binary model, finger vein recognition can be realized, which can overcome the influence of noise, improve the accuracy of recognition and be easy to implement in parallel. , but when external factors cause image blur and image deformation, etc., the local binary model cannot be used for good processing

Images