A phase unwrapping method for optical images based on u-net segmentation network

Abstract

The invention discloses an optical image phase unwrapping method based on a U-Net segmentation network. The present invention comprises the following steps: Step 1, using Zernike polynomials to generate phase-unwrapped optical images; Step 2, utilizing phase-unwrapped optical images to perform phase-wrapping operations to obtain phase-wrapped images; Step 3, utilizing U-Net to segment network training Model; step 4, using the trained model to unwrap the phase of the optical image. The present invention utilizes the U-Net segmentation network, and the method is highly targeted, and is mainly aimed at phase unwrapping of optical images. It has a great application prospect in the field of optical imaging research, and the phase unwrapping method based on the U-Net segmentation network proposed by the present invention has a high pixel segmentation accuracy rate, so that the solution accuracy rate is also high. At the same time, in the environment of CPU i5-8600K and GPU NVIDIA TITAN Xp, the solution time is only 0.037 seconds.

Description

technical field

[0001] The invention belongs to the field of image phase unwrapping, especially for optical images, and in particular relates to an optical image phase unwrapping method based on a U-Net segmentation network. Background technique

[0002] In many cases, the signal obtained by optical imaging is in complex form, including amplitude value and phase value. However, when the real phase is extracted from the complex signal, the phase value will be limited to the interval [-π, π], and the real phase outside this interval will be wound into this interval. This phenomenon is called phase winding, and the resulting phase is called a wrapped phase. Deriving the true phase from the wrapped phase is called phase unwrapping.

[0003] There are mainly three types of existing phase unwrapping methods. The first method is branch tangent method based on discrete particle swarm optimization algorithm. This method first divides the residuals of the entire image into several...

Examples

Embodiment

[0060] The training set used in the embodiment of the present invention is an optical image generated by Zernike polynomials similar to those shown in Fig. 1(a), Fig. 1(b) and Fig. 1(c). Then follow step 2 to generate the corresponding phase-wrapped images. After the model training in step 3, use the learned network for testing. The final test results are shown in Figure 3(a), Figure 3(b), Figure 3(c) and Figure 3(d).

technical field

[0001] The invention belongs to the field of image phase unwrapping, especially for optical images, and in particular relates to an optical image phase unwrapping method based on a U-Net segmentation network. Background technique

[0002] In many cases, the signal obtained by optical imaging is in complex form, including amplitude value and phase value. However, when the real phase is extracted from the complex signal, the phase value will be limited to the interval [-π, π], and the real phase outside this interval will be wound into this interval. This phenomenon is called phase winding, and the resulting phase is called a wrapped phase. Deriving the true phase from the wrapped phase is called phase unwrapping.

[0003] There are mainly three types of existing phase unwrapping methods. The first method is branch tangent method based on discrete particle swarm optimization algorithm. This method first divides the residuals of the entire image into several...

Examples

Embodiment

[0060] The training set used in the embodiment of the present invention is an optical image generated by Zernike polynomials similar to those shown in Fig. 1(a), Fig. 1(b) and Fig. 1(c). Then follow step 2 to generate the corresponding phase-wrapped images. After the model training in step 3, use the learned network for testing. The final test results are shown in Figure 3(a), Figure 3(b), Figure 3(c) and Figure 3(d).

Images