Streak image denoising and segmentation using adaptive Gaussian guided filter

Area of Science:

• * Optics and Photonics
• * Image Processing
• * Lidar Technology

Background:

• * Streak Tube Imaging Lidar (STIL) captures streak images using CCD cameras, essential for 3D imaging.
• * Image noise significantly degrades the quality of reconstructed 3D contrast and range data.
• * Effective streak image denoising requires noise reduction without sacrificing fine details.

Purpose of the Study:

• * To propose a novel adaptive Gaussian guided filter (AGGF) for streak image denoising.
• * To enhance noise removal and detail preservation in STIL images.
• * To improve the quality of 3D reconstructed images from STIL data.

Main Methods:

• * Development of an Adaptive Gaussian Guided Filter (AGGF) algorithm.
• * Integration of Guided Filter (GF) principles with Adaptive Bilateral Filter (ABF) components.
• * Optimization of the offset parameter within AGGF for enhanced detail preservation.
• * Implementation using a fast, linear-time recursive Gaussian filter kernel.

Main Results:

• * AGGF significantly sharpens denoised streak images compared to standard GF.
• * The algorithm effectively preserves image edges and thin structures.
• * AGGF outperforms existing bilateral and domain transform filters in visual quality and Peak Signal-to-Noise Ratio (PSNR).

Conclusions:

• * The proposed AGGF is a highly effective method for denoising STIL streak images.
• * AGGF achieves superior noise reduction while maintaining critical image details.
• * This method offers improved performance for 3D reconstruction in STIL applications.