Comparing CSI and PCA in Amalgamation with JPEG for Spectral Image Compression

Muhammad Safdar, Ming Ronnier Luo, Xiaoyu Liu

Abstract

Continuing previous research on color image compression, this study moves towards spectral image compression. To manage the enormous amount of data generated by spectral imaging, researchers investigated techniques to conserve image quality while improving compressibility. Principal Component Analysis (PCA) can reduce spectral dimensions for high performance, but its processing complexity is significant. As an alternative, Cubic Spline Interpolation (CSI) was considered to reduce dimensionality in the spectral domain. Both CSI and PCA were amalgamated with JPEG spatial compression. Performance was analyzed using compression rate (CR), processing time ($T_p$), and the CIEDE2000 color difference ($\Delta E_{00}$). Results show that while CSI performs slightly worse than PCA in color fidelity, it remains reliable and offers significantly lower computational complexity, especially for large-scale spectral images.

Introduction: The Spectral Data Bottleneck

Spectral imaging captures information across many more bands than standard RGB, providing a wealth of data for satellite remote sensing and medical diagnostics. However, this creates a massive storage and transmission bottleneck. To manage this, researchers use "amalgamated" compression—reducing data in both the spatial domain (the 2D layout of pixels) and the spectral domain (the "stack" of color bands).

The gold standard for spectral reduction has long been Principal Component Analysis (PCA), which identifies the most important "ingredients" of a spectrum. However, PCA is computationally expensive and data-dependent. This paper investigates whether a simpler mathematical shortcut—Cubic Spline Interpolation (CSI)—can replace PCA without sacrificing visual quality.

Methodology

The researchers compared two hybrid compression pipelines:

  1. PCA + JPEG: PCA reduces spectral dimensionality, and standard JPEG handles spatial compression.
  2. CSI + JPEG: CSI interpolates and reduces spectral bands, followed by JPEG for spatial compression.

While PCA is a standard "heavy lifter," CSI is proposed as a lightweight alternative. CSI is a polynomial-based interpolation technique that ensures smoothness at data points. Unlike PCA, it does not require complex matrix training for every image, making it computationally simple and robust. To ensure a fair comparison, the standard JPEG down-sampling stage was skipped to prevent data loss before evaluation.

Experimental Setup and Datasets

The study utilized four specific spectral test images designed to cover a range of textures and colors:

  • Caucasian Male: For testing human skin tones.
  • Red Rose & Tomatoes: For natural objects and vibrant saturations.
  • Macbeth Color Checker: A standard industry chart used to measure color accuracy across a known gamut.
Figure 3: The four spectral test images used in the study: Caucasian Male, Macbeth Color Checker, Red Rose, and Tomatoes.

Key Insights and Discussion

  • The "Tolerable" Threshold: In color science, a color difference ($\Delta E_{00}$) of less than 1.0 is generally considered invisible or tolerable to the human eye. The study found that while CSI was less precise than PCA, it consistently stayed below this 1.0 threshold.
  • The Complexity Wall: As image size increases, the processing time for PCA spikes rapidly. CSI, however, maintains a much flatter trajectory, making it a more scalable solution for high-resolution datasets.
  • Reliability over Perfection: For large-scale applications like satellite imagery, the massive gains in processing speed and lower complexity may outweigh marginal losses in color precision, provided the error remains within the "tolerable" range.

Results

Performance was measured using a fixed compression rate of 8:1.

  • Color Accuracy: PCA achieved very high accuracy with $\Delta E_{00}$ values between 0.2 and 0.3. CSI was less accurate but still reliable, with values between 0.5 and 0.8.
  • Processing Speed: The processing time ($T_p$) for PCA increased exponentially with image size. In contrast, CSI added negligible computational complexity as the number of pixels grew.
Figure 6: Comparison of processing time (ln(Tp)) between CSI and PCA as image size increases.

Limitations and Future Work

CSI is objectively "poorer" in terms of raw color difference compared to PCA. While reliable for human vision, it might not be suitable for scientific applications requiring sub-pixel spectral accuracy, such as chemical signature identification. Future research could explore adaptive compression—switching between CSI and PCA based on hardware resources or resolution—and testing these techniques at higher compression ratios (e.g., 50:1).

Summary updated

Abstract

Continuing our previous research on color image compression, we move towards spectral image compression. This enormous amount of data needs more space to store and more time to transmit. To manage this sheer amount of data, researchers have investigated different techniques so that image quality can be conserved and compressibility can be improved. The principle component analysis (PCA) can be employed to reduce the dimensions of spectral images to achieve high compressibility and performance. Due to processing complexity of PCA, a simple interpolation technique called cubic spline interpolation (CSI) was considered to reduce the dimensionality of spectral domain of spectral images. The CSI and PCA were employed one by one in the spectral domain and were amalgamated with the JPEG, which was employed in spatial domain. Three measures including compression rate (CR), processing time (Tp) and color difference CIEDE2000 were used for performance analysis. Test results showed that for a fixed value of compression rate, CSI based algorithm performed poor in terms of dE00, in comparison with PCA, but is still reliable because of small color difference. On the other hand it has lower complexity and is computationally much better as compared to PCA based algorithm, especially for spectral images with large size.
BibTeX
@misc{safdar2014comparing,
  author = {Muhammad Safdar and Ming Ronnier Luo and Xiaoyu Liu},
  title = {{Comparing CSI and PCA in Amalgamation with JPEG for Spectral Image Compression}},
  year = {2014},
  eprint = {1410.5092},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1410.5092},
}

Loading related content...

Figure