Image coding using depth blurring for aesthetically acceptable distortion.

Area of Science:

• Computer Vision
• Image Processing
• Digital Signal Processing

Background:

• Image compression techniques are crucial for efficient data transmission and storage.
• Reducing bitrate often leads to visual artifacts and loss of image quality.
• Space-variant prefiltering methods, like foveation filtering, aim to minimize perceived distortion during compression.

Purpose of the Study:

• To introduce a novel depth-based blurring prefiltering technique for image coding.
• To enhance aesthetic quality during bitrate reduction.
• To address the challenge of intensity leakage at object boundaries in space-variant blurring.

Main Methods:

• Developed a depth-based blurring algorithm that mimics camera depth-of-field effects.
• Introduced a selective blurring algorithm to simulate natural occlusion effects and prevent intensity leakage.
• Algorithm designed to handle variable blurring and occlusion levels.

Main Results:

• The proposed depth-based blurring effectively reduces high-frequency components.
• Selective blurring successfully mitigates intensity leakage at object boundaries.
• Subjective experiments demonstrated superior performance compared to foveation filtering for bitrate reduction.

Conclusions:

• Depth-based blurring offers an aesthetically superior approach to image compression.
• The selective blurring component effectively handles complex visual scenarios.
• This technique presents a promising alternative to current space-variant prefiltering methods.