A novel psychovisual threshold on large DCT for image compression

Area of Science:

• Computer Vision
• Image Processing
• Human-Computer Interaction

Background:

• Image compression relies on quantization to reduce data, but optimal values are challenging to determine.
• The human visual system's (HVS) tolerance varies with frequency, impacting perceived image quality.
• Existing methods like JPEG may not fully leverage HVS characteristics for optimal compression.

Purpose of the Study:

• To propose a psychovisual threshold for large Discrete Cosine Transform (DCT) image blocks.
• To automatically generate optimal quantization tables based on the psychovisual threshold.
• To enhance image quality and reduce artifacts in compressed images.

Main Methods:

• Developed a psychovisual threshold model based on HVS sensitivity across frequency orders.
• Applied the threshold to large DCT image blocks to derive quantization values.
• Conducted experiments comparing the proposed method with JPEG compression.

Main Results:

• The proposed psychovisual threshold effectively prescribes quantization values for each frequency order.
• Images compressed using the new method exhibit significantly fewer artifacts.
• The psychovisual threshold approach yields superior image quality at higher compression rates than standard JPEG.

Conclusions:

• A psychovisual threshold offers an effective strategy for optimizing quantization in image compression.
• This method significantly improves visual output quality and artifact reduction.
• The proposed technique outperforms JPEG compression, particularly at high compression ratios.