Decomposed Multilateral Filtering for Accelerating Filtering with Multiple Guidance Images.

Area of Science:

• Computer Vision
• Image Processing
• Signal Processing

Background:

• Multimodal signal processing and sensor fusion are critical in modern image sensing.
• Edge-preserving filtering is vital for applications like scene property estimation and inverse rendering.
• Existing accelerated filters struggle to effectively utilize multiple guidance images.

Purpose of the Study:

• To develop an efficient edge-preserving filtering algorithm capable of handling multiple guidance images.
• To extend existing efficient filtering methods to support multilateral filtering.
• To address the computational time limitations of traditional edge-preserving filters.

Main Methods:

• Proposed an algorithm named Decompose Multilateral Filtering (DMF).
• DMF decomposes the filtering process into a series of constant-time operations.
• Extends efficient edge-preserving filters to incorporate multiple guidance images.

Main Results:

• The Decompose Multilateral Filtering (DMF) algorithm demonstrates high efficiency.
• The method effectively utilizes multiple guidance images for improved filtering.
• Experimental results confirm the algorithm's suitability for diverse applications.

Conclusions:

• The proposed Decompose Multilateral Filtering (DMF) algorithm offers an efficient solution for edge-preserving filtering with multiple guidance images.
• This advancement is beneficial for various sensor fusion applications requiring enhanced image processing.
• The algorithm successfully overcomes the limitations of previous methods in handling multiple guidance data.